Georgia State University

ScholarWorks @ Georgia State University

Economics Dissertations

Department of Economics

8-11-2015

Three Essays on the Impact of the Affordable Care

Act Expansion of Dependent Coverage for Young
Adults
Yanling Qi

Follow this and additional works at: http://scholarworks.gsu.edu/econ_diss

Recommended Citation
Qi, Yanling, "Three Essays on the Impact of the Affordable Care Act Expansion of Dependent Coverage for Young Adults."
Dissertation, Georgia State University, 2015.
http://scholarworks.gsu.edu/econ_diss/116

This Dissertation is brought to you for free and open access by the Department of Economics at ScholarWorks @ Georgia State University. It has been
accepted for inclusion in Economics Dissertations by an authorized administrator of ScholarWorks @ Georgia State University. For more information,
please contact scholarworks@gsu.edu.

ABSTRACT
THREE ESSAYS ON THE IMPACT OF THE AFFORDABLE CARE ACT EXPANSION OF
DEPENDENT COVERAGE FOR YOUNG ADULTS
By
YANLING QI
AUGUST, 2015
Committee Chair: Dr. James H. Marton
Major Department: Economics
To achieve the goal of universal coverage of health insurance for the Americans, in
March 2010, the Patient Protection and Affordable Care Act (ACA) was signed into law. The
ACA targets at providing help to improve access to affordable health coverage for everyone and
protect consumers from abusive insurance company practices. One of the precedent mandates,
implemented in September 2010, is to expand coverage on young adults of age 19 to 26, who
may lose insurance coverage due to the remove from their parents plan after age 18 and lacking
of productivity to bargain with employers in the labor market.
This dissertation looks into the impact of the ACA health insurance coverage expansion
for young adults on the subsequent health outcomes, health care utilization, and further social
impact on traffic fatalities. Difference-in-differences models are used with different treatment
groups and corresponding control groups. Chapter I uses survey data (BRFSS) to evaluate health
care access, health behavior and self-assessed health status. The results suggest an improvement
in health care access and self-assessed health but more risky behavior. Chapter II uses hospital
discharge data (NIS) to estimate avoidable hospitalization in order to assess primary care
utilization. The result shows that less primary care was consumed, which leads to more avoidable
hospitalization but health may have been improved by using more hospital care. The results from
both chapters imply potential ex ante moral hazard among young adults in the policy targeting
age group. Thus, chapter III uses accident records data (FARS) to examine the impact of the
health insurance expansion on traffic fatality for young adults, to see whether young drivers
perform ex ante moral hazard through risky behavior like drunk and/or reckless driving after they
get covered by the health insurance expansion policy. Primary result shows that there is an
increase in traffic accidents and fatalities for those younger adults as a result of the ACA
dependent coverage expansion.

THREE ESSAYS ON THE IMPACT OF THE AFFORDABLE CARE ACT EXPANSION OF

DEPENDENT COVERAGE FOR YOUNG ADULTS
BY
YANLING QI

A Dissertation Submitted in Partial Fulfillment

of the Requirements for the Degree
of
Doctor of Philosophy
in the
Andrew Young School of Policy Studies
of
Georgia State University

GEORGIA STATE UNIVERSITY

2015

Copyright by
Yanling Qi
2015

ACCEPTANCE
This dissertation was prepared under the direction of the candidates Dissertation
Committee. It has been approved and accepted by all members of that committee, and it has been
accepted in partial fulfillment of the requirements for the degree of Doctor of Philosophy in
Economics in the Andrew Young School of Policy Studies of Georgia State University.

Electronic Version Approved:

Mary Beth Walker, Dean
Andrew Young School of Policy Studies
Georgia State University
August, 2015

Dissertation Chair:

Dr. James H. Marton

Committee:

Dr. Charles J. Courtemanche

Dr. M. Melinda Pitts
Dr. Erdal Tekin

DEDICATION

To Dr. James H. Marton for his unwavering support and encouragement over the years.

iv

ACKNOWLEDGEMENTS
I am profoundly grateful to a great number of people for helping me bring this work to
fruition. Thanks to these people, my graduate experience has been one that I will cherish forever.
My deepest gratitude is to my advisor, Dr. James Marton, whom without I would not be
where I am today. Any gratitude will be pale in comparison to what he dedicated to me. Thanks
for giving me outstanding guidance, expert advice, innumerable experience, extraordinary
patience and steadily support. I have been amazingly fortunate to have such an advisor who gave
me the freedom to explore on my own, led my research back on track when it faltered, and used
his wisdom to help me overcome the crisis and finish the dissertation. I hope that one day I
would become as good an advisor to my students as he has been to me.
To the rest of my committee, I cannot thank you enough for all the support and
perseverance that allowed me to successfully reach this milestone. I am especially thankful to Dr.
Charles Courtemanche for his great guidance, sincerely friendship, unwavering support and
understanding. Thanks for making me discover my potential for research, teaching me the skills
of the trade, and sharing in my joys and achievements. I am grateful as well to Dr. Melinda Pitts,
who always supported me with candid guidance and insightful suggestions. Finally, I am grateful
for insightful and detailed comments and feedback from Dr. Erdal Tekin.
I also wish to acknowledge the encouragement from Dr. Yongsheng Xu, Dr. Jon Rork
and Dr. Spencer Banzhaf for seeing my passion in research and giving me strength when I faced
challenges early in my graduate studies.
A warm gratitude to the Economists at the Federal Reserve Bank of Atlanta: Chris, Pedro,
Timothy, Federico, Lei, William, Karen, Toni, and most importantly, my supervisor Dr. Julie
Hotchkiss, who provided invaluable help on my research during my two-year stay at the bank.
Many thanks to my colleagues and friends: Subha, Ecky, Zack, Fernando, Ryan, Hal,
Marietou, Melissa, Elena, Lily, and many others who always share their happiness and
experiences with me through my stay at GSU; I am especially grateful to Vladymir and Jinjing,
who steadfastly stood by my side, encouraging me and walking me through a tough job market.
One a very personal note is to my beloved parents and family, none of this would have
been possible without the unconditional love from them, who have always supported me with
great care, faith and concern. Thanks very much for helping me make my dream of being a
scholar come true. Thanks from the bottom of my heart for a harmonious family with
warmth, peace, and love.
v

TABLE OF CONTENTS
DEDICATION ........................................................................................................................ iv
ACKNOWLDGEMENTS.........................................................................................................v
LIST OF TABLES ................................................................................................................. vii
LIST OF FIGURES .............................................................................................................. viii
INTRODUCTION .....................................................................................................................1
CHAPTER I: Impacts of the ACA Dependent Coverage Provision on Health-Related
Outcomes of Young Adults .......................................................................................................2
I. Introduction .............................................................................................................................2
II. Health Insurance and Health-Related Outcomes......................................................................7
III. Data .................................................................................................................................... 10
IV. Average Effects of the ACA Dependent Coverage Mandate ................................................ 16
V. Placebo Tests........................................................................................................................ 23
VI. Heterogeneity...................................................................................................................... 24
VII. Discussion ......................................................................................................................... 27
Tables and Figures .................................................................................................................... 33
CHAPTER II: Health Insurance and Young Adults Avoidable Hospitalizations ............... 43
I. Introduction ........................................................................................................................... 43
II. Literature Review ................................................................................................................. 46
III. Conceptual Model ............................................................................................................... 50
IV. Methodology ....................................................................................................................... 53
V. Data...................................................................................................................................... 56
VI. Results ................................................................................................................................ 60
VII. Discussion ......................................................................................................................... 64
VIII. Conclusion ....................................................................................................................... 67
Tables and Figures .................................................................................................................... 69
CHAPTER III: Health Insurance and Traffic Fatalities for Young Adults ......................... 77
I. Introduction ........................................................................................................................... 77
II. Literature Review ................................................................................................................. 79
III. Data .................................................................................................................................... 82
IV. Methodology ....................................................................................................................... 84
V. Results ................................................................................................................................. 87
VI. Discussion........................................................................................................................... 88
Tables and Figures .................................................................................................................... 90
REFERENCES ........................................................................................................................ 96
VITA ...................................................................................................................................... 104
vi

LIST OF TABLES
Table 1.1: Sample Sizes for Different Outcomes ....................................................................... 33
Table 1.2: Pre-Treatment Means and Standard Deviations for Control Variables ....................... 34
Table 1.3: Means and Standard Deviations for Outcome Variables............................................ 35
Table 1.4: Difference-in-Difference Regression Estimates of Effects of ACA Dependent
Coverage Mandate .................................................................................................................... 36
Table 1.5: Placebo Regressions ................................................................................................. 37
Table 1.6: Heterogeneity by Sex and Education......................................................................... 38
Table 1.7: Full Regression Output for Selected Dependent Variables ........................................ 39
Table 2.1: Means and Standard Deviations for Outcome Variables............................................ 69
Table 2.2: Pre-reform Means and Standard Deviations for Control Variables ............................ 70
Table 2.3: Difference-in-Differences Estimates of Effects of ACA Dependent Coverage Mandate
on Quality Indicators ................................................................................................................. 71
Table 2.4: Placebo Regressions ................................................................................................. 72
Table 2.5: Robustness Checks ................................................................................................... 73
Table 2.6: Heterogeneity by Gender and Race ........................................................................... 74
Table 2.7: Heterogeneity by Patients Zip Code Income Quartile .............................................. 75
Table 3.1: Unadjusted Difference-in-Differences Estimates of the Impact of the ACA Dependent
Coverage Expansion on Traffic Accidents and Fatalities ........................................................... 90
Table 3.2: Multivariate Difference-in-Differences Estimates of the Impact of the ACA
Dependent Coverage Expansion on Traffic Accidents and Fatalities.......................................... 91

vii

LIST OF FIGURES
Figure 1.1: Trends in Access to Care and Preventive Care Variables by Age Group .................. 40
Figure 1.2: Trends in Health Behavior Variables by Age Group ................................................ 41
Figure 1.3: Trends in Self-Assessed Health Variables by Age Group......................................... 42
Figure 2.1: Trends in Prevention Quality Indicators by Age Group............................................ 76
Figure 3.1: Traffic Accident / Fatality Counts for Young Adults Aged 20 VS 18....................... 92
Figure 3.2: Alcohol-Related Traffic Accident / Fatality Counts for Young Adults Aged 20 VS 18
................................................................................................................................................. 92
Figure 3.3: Traffic Accident / Fatality Counts for Young Adults Aged 20 VS 17-18 ................. 93
Figure 3.4: Alcohol-Related Traffic Accident / Fatality Counts for Young Adults Aged 20 VS 17-18
................................................................................................................................................. 93
Figure 3.5: Traffic Accident / Fatality Counts for Young Adults Aged 20-21 VS 17-18 ............ 94
Figure 3.6: Alcohol-Related Traffic Accident / Fatality Counts for Young Adults Aged 20-21 VS
17-18 ........................................................................................................................................ 94
Figure 3.7: Traffic Accident / Fatality Counts for Young Adults Aged 20-22 VS 16-18 ............ 95
Figure 3.8: Alcohol-Related Traffic Accident / Fatality Counts for Young Adults Aged 20-22 VS
16-18 ........................................................................................................................................ 95

viii

INTRODUCTION
To achieve the goal of universal coverage of health insurance for the Americans, in
March 2010, the Patient Protection and Affordable Care Act (ACA) was signed into law. The
ACA targets at providing help to improve access to affordable health coverage for everyone and
protect consumers from abusive insurance company practices. One of the precedent mandates,
implemented in September 2010, is to expand coverage on young adults of age 19 to 26, who
may lose insurance coverage due to the remove from their parents plan after age 18 and lacking
of productivity to bargain with employers in the labor market. The expansion shows an increase
of 3.1 million in coverage for young adults by December 2011.
This dissertation looks into the impact of the ACA health insurance coverage for young
adults on the subsequent health outcomes, health care utilization, and further social impact on
traffic fatalities. The first essay uses survey data to evaluate health care access, health behavior
and self-assessed health status. The results suggest an improvement in health care access and
self-assessed health but more risky behavior. The second essay uses hospital discharge data to
estimate avoidable hospitalization in order to assess primary care utilization. The primary result
shows that less primary care was consumed, which leads to more avoidable hospitalization but
health may have been improved by using more hospital care. The results from both essays imply
potential ex ante moral hazard among young adults in the policy targeting age group. Thus, the
third essay uses accident records data to examine the impact of the health insurance expansion on
traffic fatality for young adults, to see whether young drivers perform ex ante moral hazard
through risky behavior like drunk and/or reckless driving after they get covered by the health
insurance expansion policy. Primary result shows that there is an increase in traffic accidents and
fatalities for those younger adults as a result of the ACA dependent coverage expansion.

CHAPTER I
Impacts of the ACA Dependent Coverage Provision on Health-Related Outcomes of Young
Adults1
I. Introduction
The Patient Protection and Affordable Care Act (ACA) of March 2010 aimed to achieve
nearly universal coverage in the United States through a combination of mandates, subsidies,
Medicaid expansions, and health insurance exchanges (Gruber, 2011). Although the majority of
the ACAs provisions just took effect in 2014, one important component of the law a
dependent coverage provision was implemented on September 23rd, 2010. This provision
allows dependents to remain on a parents private health insurance plan until the start of the first
plan year after they turn 26 years old. Previously, private insurers often dropped non-student
dependents at age 19 and student dependents at age 23 (Anderson et al., 2012 and 2014).
Many states already had some form of dependent coverage mandate before the ACA, but
the state laws are typically weaker. Most state laws have an age threshold below 26 or require
additional criteria, such as being a full-time student, living with ones parents, or not being
married. Moreover, state laws do not apply to self-funded benefit programs, and more than half
of private sector workers with employer-provided health insurance are in self-funded plans
(Monheit et al., 2011). Perhaps because of these limitations, Monheit et al. (2011) and Levine et
al. (2011) find that state dependent coverage mandates only lead to small increases in dependent
coverage that are offset by a decline in young adults holding their own policies. In contrast, the
ACA provision applies to all young adults under age 26 and all private plans. It therefore has the

This chapter is coauthored with Silvia Barbaresco and Charles Courtemanche. Reprinted from Journal of Health
Economics, 40, Impacts of the Affordable Care Act Dependent Coverage Provision on Health-Related Outcomes of
Young Adults, 54-68, Copyright (2015), with permission from Elsevier.

potential to dramatically affect young adults across the country, including in states with a preexisting dependent coverage provision.
The ACA dependent coverage expansion provides a unique opportunity to study the
impacts of a health insurance intervention specific to young adults, the age group with the
highest uninsured rate (Levine et al., 2011). Prior to the ACA, the uninsured rate was 29%
among individuals ages 18-24 and 27% among those 25-34, compared to 19% for 35-44 year
olds and 14% for 45-64 year olds (DeNavas-Walt et al., 2010). Since any attempt to obtain
universal coverage necessarily involves large coverage expansions among young adults, it is
important to understand the effects of insurance on this group. It is unclear the extent to which
results from other contexts such as Medicaid, Medicare, or the Massachusetts health care
reform of 2006 are applicable. Young adults are generally healthier than the populations
covered by these programs, and therefore may experience smaller gains from health insurance.
Alternatively, young adults may be relatively poor and therefore respond strongly to reduced outof-pocket costs of medical care.2
Given the short amount of time since its implementation, researchers are only beginning
to study the impacts of the ACA dependent coverage provision. Cantor et al. (2012) and
Sommers and Kronick (2012) show that the mandate increased health insurance coverage for
young adults across all racial groups and regardless of employment status. Sommers et al. (2013)
find that the provision increased insurance coverage among young adults, while reducing delays
in getting care and care foregone because of cost. Akosa Antwi et al. (2013) again find an
increase in insurance coverage, but they also present evidence of labor market consequences
such as young adults shifting from full-time to part-time jobs. Akosa Antwi et al. (2014) show
2

Aside from age, the ACA dependent coverage mandate is also a unique coverage expansion in that it represents an
expansion of private rather than public insurance, and that, since it only affects those whose parents have insurance,
the treated population may be of higher socioeconomic status than that of other interventions.

that the mandate increased young adults utilization of inpatient care, particularly for mental
illness. Chua and Sommers (2014) do not find any evidence that the provision affected health
care use, but they do find a reduction in out-of-pocket medical expenses and increases in
excellent self-reported physical and mental health.
These papers all share a common general research design: comparing changes in
outcomes among the treated age range 19-25 to those of other young adults. The age range used
for the control group varies across these studies, with some including individuals up to 34 years
old (Sommers and Kronick, 2012; Sommers et al., 2013; Chua and Sommers, 2014). Slusky
(2013) questions the validity of this approach, arguing that different age groups are often subject
to different economic shocks. He runs placebo tests using data from before the mandate and
artificial treatment dates, finding that the same specification estimates significant effects
more often than could be attributed to chance. He suggests narrowing the age bandwidths of the
treatment and control groups as a possible solution.
We contribute to this literature on the ACA dependent coverage provision in four ways.
First, we consider a number of new outcomes. Using data from the Behavioral Risk Factor
Surveillance System (BRFSS), we investigate 18 outcomes related to health care access,
utilization of preventive care, risky health behaviors, and self-assessed health. The health care
access measures include having insurance, a primary care doctor, and any foregone care because
of cost. Our preventive care measures are dummies for recent flu vaccinations, well-patient
checkups, and pap tests. The health behavior outcomes reflect smoking, drinking, body mass
index, exercise, and pregnancy. The self-assessed health variables relate to overall, mental, and
physical health as well as health-related functional limitations. Of these outcomes, only
insurance coverage, foregone care because of cost, and self-assessed physical and mental health

are studied in other papers in the literature. To our knowledge we are the first to investigate the
ACA dependent coverage provisions impact on preventive care or health behaviors. Moreover,
although Chua and Sommers (2014) examine self-assessed physical and mental health, their
measures and ours are meaningfully different. They use dummies for self-reporting excellent
physical and mental health, so their estimates only capture changes at the upper end of the health
distribution. In contrast, we utilize five measures that should together capture changes at various
parts of the distribution. A dummy for excellent overall health reflects the high end, a dummy for
very good or excellent health reflects a somewhat lower portion, and three more severe outcomes
number of days of the past 30 not in good physical health, not in good mental health, and with
health-related limitations reflect an even lower portion. This distinction will prove critical to
the results.
Our second contribution is to push further than prior studies toward addressing the
methodological concerns raised by Slusky (2013), both by using narrow age ranges for the
treatment and control groups and by validating these selections through placebo testing. Our
treatment group consists of individuals ages 23-25, slightly below the dependent coverage
provisions age cutoff, and our control group consists of those slightly above the cutoff at ages
27-29. We run placebo tests checking for effects of artificial interventions in the pre-treatment
period. Our classifications perform well in the placebo tests, whereas the wider age ranges
commonly used in the literature prove more problematic.
Another contribution is that we use over three full years of post-treatment data (2011
through 2013, plus a few months after implementation at the end of 2010). To our knowledge,
none of the prior papers in the ACA dependent coverage provision literature have used more
than one full year of post-treatment data, which leaves the estimates susceptible to confounding

from temporary age-specific shocks and fluctuations. If estimated effects persist with three years
of post-treatment data, we can be more confident that they are not driven by transitory
movements in unobserved characteristics.
Finally, we contribute to the literature by testing for heterogeneous effects. Of the
outcomes included in our paper, heterogeneity in the effects of the ACA dependent coverage
provision has only previously been evaluated for insurance coverage (Akosa Antwi et al., 2013;
Sommers et al., 2013) and cost being a barrier to care (Sommers et al., 2013). We will find
important heterogeneous effects on other outcomes as well, such as self-assessed health.
Moreover, although Akosa Antwi et al. (2013) and Sommers et al. (2013) evaluate whether
effects differ by certain demographic characteristics, neither paper tests for heterogeneous effects
by socioeconomic status.3 We will find that the effects of the dependent coverage provision vary
considerably by education level.
Our difference-in-differences results from the full sample suggest that the ACA
dependent coverage provision improved health care access for young adults, had little effect on
preventive care use, had mixed effects on risky health behaviors, and improved self-assessed
health at the high end of the distribution. Specifically, we document improvements in four of the
eighteen outcomes: health insurance coverage, access to a primary care doctor, excellent selfassessed health, and body mass index. However, we find evidence of an increase in risky
drinking, and no clear effects in either direction on the remaining thirteen outcomes.
We evaluate heterogeneity in the effects of the mandate through subsample analyses,
finding the greatest improvements in outcomes for men and college graduates. The increase in

Sommers et al. (2013) note that testing for heterogeneity by educational attainment is difficult because many
individuals in their treatment group 19 to 25 year olds are still in the process of completing their education.
Another advantage of using a narrow age range for the treatment group 23 to 25 year olds is that excluding the
prime college ages largely ameliorates this concern.

health insurance coverage was greater for men than women, and only men experienced
statistically significant gains in any outcomes beyond health insurance: primary care access,
exercise, and overall self-assessed health. Stratifying by education reveals that the insurance
expansions were similar for college graduates and non-college graduates. However, only college
graduates experienced significant gains in any other outcomes besides insurance specifically,
primary care access, cost being a barrier to care, body mass index (BMI), obesity, and overall
self-assessed health. Young adults with different education levels therefore appear to respond
differently to exogenously obtaining health insurance.
II. Health Insurance and Health-Related Outcomes
The most obvious theoretical implication of health insurance is that by lowering the
effective price of health care, health insurance should increase its utilization. However, increased
health care utilization does not necessarily improve health. Diminishing marginal returns suggest
that health care can only improve health up to a certain level (e.g. Grossman, 1972). Whether the
additional consumption of medical care induced by insurance generates substantial gains in
health therefore depends on the initial level of health capital. Since the uninsured can often
obtain essential needs by paying directly or receiving charity care, these individuals need not
have low baseline levels of health. Moreover, the marginal returns to health care differ for
different outcomes. Risky health behaviors such as smoking, excessive drinking, and overeating
might be particularly difficult to improve through health care, as they require lifestyle changes.
Medical professionals ability to influence health behaviors is generally limited to providing
accountability, information, strategies, and sometimes drugs to make behavioral changes easier.
Another relevant issue when evaluating the impact of health insurance on health is that
obtaining insurance could induce individuals to take more health risks, since the provision of

health insurance decreases the financial losses associated with sickness. This concept is known
as ex ante moral hazard (Ehrlick and Becker, 1972). Theoretically, ex ante moral hazard could
both increase risky behaviors and reduce investments in preventive care.
Finally, exogenous provision of health insurance could lead to income effects for
individuals who used to purchase their own insurance policy but now are able to receive free or
subsidized coverage, or for the newly-insured if their out-of-pocket medical expenses drop. The
available evidence from natural experiments suggests that additional income increases health
care utilization (Acemoglu et al., 2013), either increases BMI or has no effect (Lindahl, 2005;
Schmeiser, 2009; Cawley et al., 2010), increases smoking along the intensive but not extensive
margin (Apouey and Clark, 2014), and increases drinking (Apouey and Clark, 2014). The
income effect may therefore improve health via medical care but worsen health via risky
behaviors. Accordingly, evidence of incomes causal effect on overall health is mixed, with
Lindahl (2005) and Frijters et al. (2005) finding that it improves self-assessed health, Apouey
and Clark (2014) finding that it improves mental health but not overall health, and Snyder and
Evans (2006) showing that it raises mortality risk among seniors.
In sum, the effects of insurance on preventive health care utilization, risky health
behaviors, and overall health status are theoretically ambiguous. Insurance may improve these
outcomes through direct price effects, worsen them through ex ante moral hazard, or affect them
in either direction through income effects. The net effects could differ for different outcomes.
For instance, direct price effects might dominate for primary care utilization but moral hazard
might dominate for risky behaviors. Empirical analysis is necessary to resolve this ambiguity.
Causally interpretable evidence generally confirms the prediction that insurance increases
health care utilization for U.S. adults. Manning et al. (1987) analyzed the randomized RAND

Health Insurance Experiment, finding that lower copayments increased doctor visits. Medicaid
and Medicare expansions have been shown to increase utilization of primary and hospital care
(Currie and Gruber, 1996a; Finkelstein et al., 2012; Taubman et al., 2014; Lichtenberg, 2002;
Card et al., 2008). Other evidence suggests that the Massachusetts universal coverage initiative
of 2006 increased preventive services while reducing emergency room utilization, avoidable
hospitalizations, and medical needs unmet because of cost (Miller, 2011; Kolstad and Kowalski,
2012; Miller, 2012; Van der Wees et al., 2013). More directly relevant to our study population,
Anderson et al. (2012 and 2014) exploit the sharp drops in coverage on parents insurance at
ages 19 and 23 to show that losing coverage reduced young adults emergency room and hospital
visits. Finally, as mentioned previously, Akosa Antwi et al. (2014) show that the ACA dependent
coverage provision increased hospital admissions, although Chua and Sommers (2014) find no
significant effects on survey measures of hospital, primary care, or prescription drug utilization.
The evidence of health insurances effect on health is mixed. The RAND experiment only
found that better insurance coverage improved health for certain subgroups (Brook et al., 1983).
Medicaid expansions increase self-reported overall, physical, and mental health and reduce
mortality, but have no statistically detectable effects on laboratory-measured health outcomes
(Currie and Gruber, 1996b; Finkelstein et al., 2012; Sommers et al., 2012; Baicker et al., 2013).
Card et al. (2009) find a reduction in the mortality rate among recently hospitalized Medicare
recipients, but Finkelstein and McKnight (2008) find no significant effect of Medicare on the
mortality rate of seniors in general. Evidence suggests that the Massachusetts reform improved
self-assessed overall, physical, and mental health, while decreasing functional limitations, joint
disorders, and mortality (Van der Wees et al., 2013; Courtemanche and Zapata, 2014; Sommers

et al., 2014). As mentioned previously, Chua and Sommers (2014) find that the ACA dependent
provision increased the probabilities of self-reporting excellent physical and mental health.
Evidence on the causal effects of health insurance on risky health behaviors is also mixed.
Brook et al. (1983) find no evidence that insurance affected smoking or body weight in the
RAND experiment. Dave and Kaestner (2009) report that Medicare decreased physical activity
while increasing smoking and drinking. Finkelstein et al. (2012) do not find any significant
impacts of Medicaid on smoking or BMI. Courtemanche and Zapata (2014) find that the
Massachusetts reform reduced body mass index and did not affect smoking or physical activity.
In sum, there is little prior evidence on the effects of health insurance on young adults
access to care, preventive care utilization, risky health behaviors, or health. Given the theoretical
ambiguities and variation in empirical findings discussed above, we cannot assume prior results
from other contexts such as Medicaid and Medicare generalize. For instance, young adults
relatively high baseline levels of health might lead them to have relatively inelastic demand for
health care or a low marginal effect of health care on health. On the other hand, young adults
demand for health care could be relatively elastic given their generally low income and wealth
levels. Moreover, one might expect young adults to be the most susceptible to ex ante moral
hazard since this is often the life stage in which opportunities to engage in particular risky
behaviors (e.g. binge drinking) are introduced.
III. Data
Our main data source is the BRFSS, a telephone survey conducted by state health
departments in conjunction with the U.S. Centers for Disease Control and Prevention to collect
information on health and health behaviors. The survey is conducted monthly through a random
digit dialing method that selects a representative sample of respondents from the non-

10

institutionalized population of adults at least 18 years old. The BRFSS provides several
advantages for our analyses. First, it contains a wide range of appropriate outcome variables.
Second, it includes demographic characteristics as well as state, month, and year identifiers that
allow us to construct the treatment variable and jointly control for many different factors. Next, it
contains a much larger number of observations than other datasets with the necessary variables.
Finally, the BRFSS includes a number of pre-treatment waves that allow for detailed testing of
differential trends in the outcomes between treatment and control groups.
Our primary analysis sample consists of the 2007-2013 waves, which include the year the
ACA dependent coverage mandate took effect plus three years on both sides. One reason we
exclude the years before 2007 is to limit our sample to years of relatively poor economic
performance. This reduces the possibility of confounding from differential impacts of
macroeconomic shocks on the health-related outcomes of different age groups. However,
robustness checks and placebo tests will utilize data as far back as 2001. We do not use any
waves before 2001 because the BRFSS made major changes to the survey in that year. Many of
the questions used to construct our outcome variables are either not available in earlier years or
differ in non-trivial ways.
Most of our analyses use ages 23-25 as the treatment group and ages 27-29 as the control
group. Following much of the prior literature, 26 year olds are excluded because their treatment
status is ambiguous: they may still be covered by the ACA mandate depending on their birthdate
and the start date of their parents insurance plan year (Akosa Antwi et al., 2013). Although the
prior literature uses 19-25 as the treatment group, we prefer 23-25 for two reasons.4 First, prior to
the ACA, insurers most commonly dropped non-student dependents from parents plans at age
4

Studies in the literature utilize somewhat different control groups. Cantor et al. (2012) use 27-30 year olds;
Sommers and Kronick (2012), Sommers et al. (2013), and Chua and Sommers (2014) use 26-34 year olds; Akosa
Antwi et al. (2013) use 16-18 and 27-29 year olds; and Akosa Antwi et al. (2014) use 27-29 year olds.

11

19, but most commonly dropped student dependents at age 23. Excluding 19-22 year olds
therefore results in a cleaner treatment group, i.e. a higher proportion of the treatment group
actually being affected by the treatment. Accordingly, Akosa Antwi et al. (2014) show that the
ACA dependent coverage provisions impact on having insurance was more than twice as large
for 23-25 year olds as for 19-22 year olds. Second, Slusky (2013) shows that the models from
prior papers with ages 19-25 as the treatment group lead to poor placebo test results for insurance
and labor market outcomes. He suggests narrowing the age bandwidth as a potential solution.
Indeed, we will show that wider age ranges lead to problematic placebo test results for our
outcomes as well, and that our narrower age range performs better.
We utilize eighteen different health-related dependent variables. The first three relate to
health care access: dummy variables reflecting whether the respondent has any health insurance,
has a primary care physician, and had any medical care needed but not obtained because of cost
in the previous year. Unfortunately, the BRFSS does not include more detailed questions on
health insurance, such as the source of coverage. The next three outcomes dummies for having
a flu vaccination (shot or spray), a well-patient doctor check-up visit (e.g. physical), and a pap
test (for women) in the previous year reflect preventive care utilization.5 The next category of
variables relates to risky health behaviors: a dummy for whether the individual currently smokes,
number of alcoholic drinks in the past 30 days, a dummy for being a risky drinker (more than 30
drinks total or at least one occasion with four or more drinks for women, more than 60 drinks
total or at least one occasion with five or more drinks for men),6 body mass index (BMI=weight

Other preventive care variables typically studied in the literature, such as mammograms and prostate exams, are
not relevant for our study population of young adults.
6
The dummy for risky drinker is created to come as close as the BRFSS data will allow to the National Institute on
Alcohol Abuse and Alcoholisms definition of at-risk drinking: more than 7 drinks per week total or at least one
occasion with three or more drinks for women, and more than 14 drinks per week total or at least one occasion with
four or more drinks for men. See http://pubs.niaaa.nih.gov/publications/womensfact/womensfact.htm.

12

in kg/height in m2),7 a dummy for obese (BMI30), a dummy for whether an unmarried female
respondent is pregnant (the only proxy for risky sexual activity available in the BRFSS), and a
dummy for obtaining any recreational exercise in the past 30 days.8 Finally, we include several
variables related to self-assessed health status: a dummy for whether overall health is very good
or excellent, a dummy for whether overall health is excellent, and days of the last 30 not in good
mental health, not in good physical health, and with health-related functional limitations.
Although self-assessed health is subjective, research has repeatedly found it to be correlated with
objective measures of health such as mortality (e.g. Idler and Benyamini, 1997; DeSalvo et al.,
2006; Phillips, Der, and Carroll, 2010). Self-assessed health is also a global measure of health
that captures the full range of possible diseases and limitations (Idler and Benyamini, 1997).9
We also utilize a wide array of control variables. These include dummy variables for each
year of age, gender, race/ethnicity, marital status, education, household income category, number
of children in the household, whether the respondent reports her primary occupation as student,
and whether the respondent is unemployed. Additionally, we control for monthly state
unemployment rate, obtained from the Bureau of Labor Statistics. As mentioned previously, we
are concerned about different impacts of the recession on different age groups, so controlling for
several variables related to economic conditions at both the individual and aggregate levels could
potentially be important. We also control for whether the respondents state had any dependent
coverage mandate covering her age*marital status*student status group in the survey year based

Body mass index is based on self-reported height and weight, which are prone to measurement error (Cawley,
2004). Researchers have repeatedly found that this measurement error does not affect the signs and significance of
regression estimates with BMI as a dependent variable, though it may slightly attenuate the magnitude of the
estimates (e.g. Lakdawalla et al., 2002; Courtemanche et al., 2014; Courtemanche et al., forthcoming).
8
Unfortunately, the more detailed BRFSS questions on physical activity are only available in odd numbered survey
years and changed dramatically in 2011, so they are not useful for our analyses.
9
Moreover, other commonly-used measures of health are not practical in our context. Mortality rates are likely too
low among young adults to estimate effects of coverage expansions with meaningful precision, while measures of
avoidable hospitalizations confound insurances impact on health with the reduction in effective prices.

13

on information from the National Conference of State Legislatures (2010).10 Additionally, in the
flu vaccination regressions we control for interactions of the age fixed effects with the number of
positive influenza tests in the country during the particular flu season (a proxy for severity of the
flu season). Flu seasons in the post-treatment years were much more severe than those in the pretreatment years, so adding these interactions prevents the estimates from being confounded by
differential responses to flu season severity by young adults of different ages.11
Finally, we include a dummy for whether the respondent is part of a cell phone only
component of the sample, added in 2011 (this variable is 0 for all respondents before 2011). The
fact that individuals who only used cell phones were not explicitly included in the sample until
2011 raises the question of whether our sample makeup meaningfully changed at about the same
time the post-treatment period began. To address this issue, we not only control for cell phone
only users but also utilize the BRFSS sampling weights in all analyses. We found that these
weights eliminate any sharp changes in sample demographic characteristics in 2011.
Additionally, this issue would only bias our regression estimates if the relationship between the
outcomes of landline and cell phone users is different among 23-25 year olds than among 27-29
year olds, and in a way that is not captured by the controls. It is not obvious why this would be
the case. Accordingly, we have verified (results available upon request) that dropping individuals
who only use cell phones from our sample has very little effect on the coefficient estimates,
though it does generally increase the standard errors due to the reduced sample size.
10

Note that not everyone coded as a 1 for state mandate is actually treated by such a mandate. Additional
qualifiers beyond age, student status, and marital status exist in some states, while young adults whose parents
employers self-insure are also not covered by state mandates.
11
Specifically, for the pre-treatment years 2007, 2008, and 2009, there were 23,753, 39,827, and 27,682 positive
influenza test results in the corresponding flu seasons 2006-2007, 2007-2008, and 2008-2009. For the post- or
during-treatment years 2010, 2011, 2012, and 2013, there were 157,449, 55,403, 27,012, and 75,342 number of
influenza test results in the corresponding flu seasons 2009-2010, 2010-2011, 2011-2012, and 2012-2013 (CDC,
2014). The large 2009-2010 flu season number largely reflects the swine flu pandemic, but two of the three
subsequent seasons were still relatively strong. Our results suggest that younger young adults respond more strongly
to flu season severity than older young adults; therefore, omitting these interactions would lead to biased estimates.

14

After excluding observations with missing data for any of the control variables, Table 1.1
reports the sample sizes for the regressions for each dependent variable, along with the numbers
of individuals in the treatment and control groups. The sample sizes differ slightly across
dependent variables for two reasons. First, each health-related variable is missing for a different
number of respondents. Second, the health-related variables have different reflection periods;
some apply to the present (e.g. current smoker), while others refer to a 30-day period (e.g.
number of alcoholic drinks in the past 30 days) and others to a one-year period (e.g. any wellpatient doctor visit in the past year). We are concerned that short-run estimates would be
misleading for variables with a long reflection period.12 We therefore drop respondents surveyed
during this period of ambiguity; e.g. for well-patient doctor visit in the past year we drop October
2010 through September 2011, while for drinks in the past 30 days we drop only October 2010.13
Table 1.2 lists the control variables and compares the pre-treatment (January 2007
through September 2010) summary statistics of the treatment and control groups. Individuals in
the treatment group are less likely to be married, have a college degree, earn a high income, and
have children in the household, and they are more likely to be students or employed.
Table 1.3 reports the pre- and post-treatment sample means of the outcome variables for
the treatment and control groups, and calculates the simple difference-in-difference of means.
Prior to the ACA dependent coverage provision, the uninsured rate was higher for young adults
in the treatment group than those in the control group. The treatment group had lower rates of
health care utilization and health care access than the control group; higher drinking and
12

For example, suppose a respondent is surveyed in November 2010, the second month of the post-implementation
period. The respondent would be classified as post-treatment, but her answer about well-patient doctor visits in the
past year would reflect only two months of the post-treatment period and ten months of the pre-treatment period.
13
For flu vaccinations in the past year, we only drop October 2010 through December 2010, as opposed to dropping
a full year. We feel a shorter reflection period is appropriate in this case because flu vaccinations are typically
administered in the fall. For instance, if someone surveyed in March 2011 reports being vaccinated in the past year,
that vaccine almost certainly occurred during the post-treatment period (October 2010 or later).

15

unmarried pregnancy rates but healthier levels of risky drinking, BMI, obesity, and exercise; and
broadly similar levels of smoking and self-assessed health. Comparing changes in the post- and
pre-treatment means for the treatment and control groups, the difference-in-differences are
positive and significant for any insurance, primary care doctor, excellent health, and risky
drinker; negative and significant for body mass index and obesity; and insignificant for the other
outcomes including all those in the preventive care category.
Simple difference-in-differences estimates account for fixed differences in unobservable
characteristics between the treatment and control group, but are still susceptible to bias from
time-varying observables and unobservables. Figures 1-3 show that at a first glance the pre-ACA
trends for the treatment and control groups appear generally similar for most outcomes,
providing preliminary evidence that changes over time in observables and unobservables may
not be substantially different for 23-25 year olds and 27-29 year olds. We next turn to regression
analyses that adjust for changes in observables. Later, we will also conduct more formal tests of
the assumption of common trends in unobservables.
IV. Average Effects of the ACA Dependent Coverage Mandate
A. Baseline Model
We estimate the effects of the ACA dependent coverage provision on the eighteen healthrelated outcomes using reduced-form difference-in-differences regressions. While it is tempting
to estimate instrumental variables models using the mandate as an instrument for having
insurance coverage, we are not confident that the exclusion restriction would hold in such
models because there are several other mechanisms through which the mandate could affect
health-related outcomes besides the extensive margin of health insurance coverage. Other

16

possible mechanisms include the intensive margin of coverage (switching from high deductible
catastrophic coverage to more comprehensive coverage), income effects, and peer effects.
Our baseline regression is of the form
=
where

(1)

is the health-related outcome for individual i of age g living in state s in time t,

expressed in a month/year combination.14

is a dummy variable for whether age g is in the

treated age range 23-25 as opposed to the control age range 27-29.
period t is after the implementation of the provision (October 2010 or later).

indicates whether
is the difference-

in-differences coefficient and it captures the difference between the effects of the mandate on the
treatment and control groups.

is a vector of the aforementioned control variables for sex,

race, marital status, education, income, children, cell phone survey, student status, individual and
state unemployment, and state dependent coverage mandate. We also include fixed effects for
each year of age, month/year of time (e.g. January of 2007), and state, denoted by
respectively.
model because

is the error term. 15 We do not separately include

is perfectly collinear with the age fixed effects while

and

, and

in the
is perfectly

collinear with the month/year fixed effects.

We report heteroskedasticity-robust standard errors clustered at the level of treatment:
age. Following convention when there are a small number of clusters (six in our case), for
hypothesis testing we use a t-distribution with degrees of freedom equal to the number of clusters
minus one. The critical values used in our hypothesis tests are therefore considerably more
14

Even though most of our outcomes are binary or non-negative count, we estimate linear models because they
typically give reliable estimates of average effects (Angrist and Pischke, 2008). In unreported regressions (available
upon request), we verify that the average treatment effects are very similar using probit regressions for the binary
outcomes and negative binomial regressions for the count outcomes.
15
In unreported regressions (available upon request) we have verified the results remain virtually identical if we
replace the state fixed effects with fixed effects for each state-by-year combination.

17

stringent than those using the standard normal distribution. It is possible that even using stringent
critical values might not be sufficient to eliminate the tendency to over-reject when the number
of clusters is small (Cameron et al., 2008). However, the placebo tests in the next section will
reject the null hypothesis even fewer than the expected number of times, suggesting that our
hypothesis tests are sufficiently conservative. One of our robustness checks will also address this
issue.
The key identifying assumption in a difference-in-differences model is common
counterfactual trends between the treatment and control groups; i.e. in the absence of the
intervention the treatment and control groups would have experienced the same changes in
outcomes. Slusky (2013) argues that this assumption is problematic when studying the impact of
the ACA dependent coverage provision on labor market-related outcomes (e.g. sources of health
insurance coverage, employment status, and work hours) since cyclical fluctuations in the
economy have different effects on different age groups. Since economic fluctuations are related
to health, 16 Sluskys concern could also apply to health-related outcomes. As discussed
previously, this is one of our main reasons for using narrow age bandwidths of 23-25 and 27-29.
B. Robustness Checks
We also estimate several variations of (1) as robustness checks. First, we run regressions
including only the demographic controls (the sex, age, race, children, and marital status dummies)
and fixed effects, excluding the economic controls since they may be endogenous to the
dependent coverage provision. Obtaining access to parents insurance could potentially influence
a young adults decisions about employment and education, which would then affect income.

16

Research generally shows that recessions are associated with improvements in health and health behaviors (e.g.
Ruhm, 2000, 2002, 2005), although recent evidence suggests that the countercyclical nature of health observed in
prior recessions may not have been present during our sample period (Ruhm, 2013; Tekin et al., 2013).

18

Including covariates related to employment, education, and income might therefore control
away part of the causal effect of the policy.
Our next several robustness checks vary the time period included in the sample. In order
to verify that the results are not driven by our chosen length of the pre-treatment period, we
consider two alternatives: starting the sample in 2004 and 2001. Additionally, we run regressions
dropping March 2010 through December 2010, as these months are somewhat ambiguous with
respect to their treatment status. We drop March-September because the ACA was passed in
March, so some insurance plans may have complied preemptively prior to the dependent
coverage provisions official implementation in September. We drop October-December because,
even though the mandate was implemented in September, insurers did not have to comply until
the start of the next plan year, which is often January.17
Our final robustness check addresses the potential concern that standard errors may be
understated because of autocorrelation given the small number of clusters. We collapse the data
into one observation for each year of age in the pre-treatment period and one observation for
each year of age in the post-treatment period, for a total of twelve observations. We then estimate
=

where the lines above variables indicate averages across all individuals of age

(2)

in time period

(pre- or post-treatment) , weighted by the individual BRFSS sampling weights. Since the small
sample size prevents all the control variables from being separately included,
variable that summarizes the influence of all the controls.

is a single

is computed by regressing outcome

on the controls using the individual-level pre-treatment data, then predicting

for the whole

sample based on the coefficient estimates, then aggregating in the same manner described above.
17

Akosa Antwi et al. (2013) include two treatment variables to separately model the effects of the mandate during
the implementation period and after full implementation. We have considered this specification in unreported
regressions and the estimated post-implementation effects remain very similar.

19

C. Results
Table 1.4 presents the results for the baseline model and robustness checks. In addition to
reporting estimated treatment effects and standard errors, for the baseline regressions we also
report (in brackets) the treatment effects expressed in standard deviations of the dependent
variables to provide some comparability of effect sizes across the different outcomes.
The results suggest sizeable improvements in health care access along at least some
dimensions. We estimate that the ACA dependent coverage provision statistically significantly
increased the insurance coverage rate of 23-25 year olds by between 5.5-6.7 percentage points,
depending on the model. This is somewhat larger than the around 3-5 percentage point increase
estimated by previous studies that use the broader treated age range of 19-25 (Cantor et al., 2012;
Sommers and Kronick, 2012; Akosa Antwi et al., 2013; Sommers et al., 2013).18 Additionally,
the mandate increased the probability of having a primary care doctor by 2.0-3.4 percentage
points and decreased the probability of having any care needed but foregone because of cost by
1.6-2.3 percentage points. The effect on primary care doctor access is statistically significant in
all specifications, but the effect on care foregone because of cost is never significant.
Despite this improved access, we do not find any evidence of increased preventive care
utilization. We estimate a total of eighteen models across the three preventive care measures, and
none of these models reveal a statistically significant positive effect of the dependent coverage
provision. The estimated effects on flu vaccinations and pap tests are negative in most
specifications and occasionally statistically significant. The estimates for well-patient checkup
are all positive but never significant.

18

This discrepancy is consistent with Akosa Antwi et al.s (2014) finding that the mandates impact on the
probability of having any coverage was around twice as large for 23-25 year olds than 19-22 year olds (4 compared
to 2 percentage points). Alternatively, estimates using the treated age range 19-25 could be biased downward given
the problems documented in our placebo tests and those of Slusky (2013).

20

We find mixed evidence regarding the dependent coverage provisions impacts on risky
health behaviors. No significant estimates are observed for smoking, pregnancy, or alcoholic
drinks per month. However, the mandate statistically significantly increased the probability of
risky drinking (excessive drinks per month or any binge drinking) in all specifications, with
magnitudes ranging from 0.8-1.4 percentage points. The dependent coverage expansion therefore
appears to affect drinking at only the high end of the distribution, which is consistent with an ex
ante moral hazard explanation since mild to moderate drinking generally does not increase the
need for medical services. In contrast, the dependent coverage provision appears to improve
weight-related behaviors. The mandate reduces BMI in all six specifications, with magnitudes
ranging from -0.098 to -0.175. All but one of the six estimates for BMI are significant, with the
remaining one being nearly significant. The effect on obesity is also negative in all six models,
though it is only significant in three. The effect on probability of having any exercise is positive
in all specifications but only significant in one. It is possible that our inability to measure
exercise in greater detail e.g. calories burned per day from physical activity prevents the
emergence of further significant results. It is also possible that the reduction in BMI is coming
via reduced caloric intake, which we are unable to measure in the BRFSS.
It is theoretically conceivable that insurance coverage could increase risky drinking but
reduce weight. Health care access may be more helpful for losing weight than reducing drinking.
Gains in information and accountability may both be greater for weight control than drinking:
dieting strategies can be complicated and benefit greatly from professional advice, and
accountability is greater for weight since patients are weighed at each visit. Additionally, the ex
ante moral hazard effect could be stronger for risky drinking than weight-related behaviors.
Binge drinking has a non-trivial chance of resulting in immediate medical needs, either from

21

alcohol poisoning, drunk driving accidents, or other injuries.19 In contrast, expenditures to treat
diseases associated with obesity typically occur years down the road. Perhaps uninsured young
adults assume that they will be insured by time these downside risks are realized, in which case
ex ante moral hazard would not apply. In short, the direct price effect could dominate for BMI,
while the ex ante moral hazard effect could dominate for drinking. Income effects may play a
role as well, especially for alcohol consumption given the aforementioned evidence of a positive
causal effect of income on drinking (Apouey and Clark, 2014).
Turning to the self-assessed health outcomes, the mandate increased the probability of
young adults reporting excellent overall health by 1.3-1.5 percentage points and very
good/excellent health by 1.1-1.8 percentage points. However, only the estimates for excellent
health are significant, as the standard errors for very good/excellent health are larger. We do not
find any evidence of effects on the variables representing more severe health problems: days not
in good mental health, not in good physical health, and with health-related functional limitations.
The lack of effects on our mental and physical health outcomes is particularly interesting in light
of Chua and Sommers (2014) finding that the ACA dependent coverage provision increased the
probabilities of reporting excellent mental and physical health. Chua and Sommers mental and
physical health variables emphasize changes at the high end of the health distribution and may
therefore correspond more closely to our variable for excellent overall health than our physical
and mental health variables, which focus on not good health. In other words, both our results
and those of Chua and Sommers are consistent with the provisions effects on mental and
physical health being concentrated in the high end of the health distribution.

19

In the US, approximately 80,000 cases of alcohol poisoning and 10,322 alcohol-impaired driving crashes occur
annually, with these incidents disproportionately involving young adults (CDC, 2012; NHTSA, 2014). 599,000
alcohol-related injuries occur annually among 18-24 year old college students (NIAAA, 2013).

22

Finally, we provide a brief discussion of the relative magnitudes of the effects on

different outcomes by comparing the treatment effects expressed in standard deviations of the
dependent variables. Not surprisingly, the largest effect of 0.13 standard deviations is on the
probability of having any health insurance coverage. The next largest statistically significant
effect is on primary care doctor access (0.065 standard deviations), then excellent health (0.032
standard deviations), then risky drinker (0.026 standard deviations), then finally BMI (-0.017
standard deviations). The largest statistically insignificant effects are on flu vaccinations (-0.033
standard deviations) and very good/excellent health (0.031 standard deviations).
V. Placebo Tests
We next provide a series of placebo tests to evaluate whether the previous results can
credibly be interpreted as causal effects of the ACA dependent coverage provision. Following
Slusky (2013), we estimate variants of equation (1) that test for effects of artificially-timed
treatments during pre-treatment years. We estimate models for three different seven-year
windows of pre-treatment data (to match the seven years used in our main 2007-2013 analyses):
2003-2009, 2002-2008, and 2001-2007. Since the first month after the implementation of the
actual dependent coverage mandate was the 46th month (October 2010) of our 2007-2013 sample,
in each placebo test sample we date the implementation of the artificial intervention to the 46th
month (e.g. October 2006 for the 2003-2009 sample). We estimate (1) for each of the eighteen
dependent variables in each of the three placebo test samples.
Table 1.5 reports the coefficient estimates of interest from these placebo tests. We run
three tests for each of the eighteen dependent variables, though a test is not possible for checkups
using 2001-2007 data since the checkup question was not asked until 2005. This leaves a total of
53 regressions. Given the large number of estimates, we would expect some significant results

23

even for valid models. Specifically, approximately 0-1 estimates should be significant at the 1%
level, about 2-3 at the 5% level, and about 5 at the 10% level. We obtain numbers even smaller
than these. No estimated treatment effects are significant at the 1% level, 2 (3.8%) are
significant at the 5% level, and 3 (5.7%) are significant at the 10% level. Moreover, we do not
obtain more than one placebo test rejection for any outcome. In other words, it is not clear that
there are any outcomes for which our baseline difference-in-differences model is inappropriate.
In the interest of contributing to the broader debate in the literature about the
appropriateness of different age bandwidths when using difference-in-differences models to
estimate the effects of the ACA dependent coverage provision, we also run the same set of
placebo tests for the most common age ranges used in the literature: treatment group 19-25 and
control group 26-34 (Sommers and Kronick, 2012; Sommers et al., 2013; and Chua and
Sommers, 2014). We obtain 4 placebo test rejections (7.5%) at the 1% level, 7 (13.2%) at the 5%
level, and 11 (20.8%) at the 10% level. The full table of results is available upon request.
VI. Heterogeneity
Having established our baseline results and assessed the validity of our model, we next
turn to an examination of heterogeneity in the treatment effects. We considered stratifications by
sex, race/ethnicity, education, and state pre-ACA dependent coverage law status, but we did not
observe any statistically significant differences in effects across the subgroups for race/ethnicity
and pre-ACA law, so we only report the results for the stratifications by sex and education. For
education, we stratify into two groups: college graduates and non-college graduates.20
Theoretically, the ACA dependent coverage provision could have heterogeneous effects
on health-related outcomes for three reasons. First, there could be heterogeneous effects on the
20

Further stratification by education led to estimates that were too imprecise to be useful. Note that we do not
include a separate category for current students because our sample only includes those 23 and older, so the
proportion of our respondents reporting student as their primary occupation is low.

24

probability of having insurance coverage. In the pre-treatment portion of sample, females were
more likely to have insurance than males (76% versus 67%), and college graduates were much
more likely to have insurance than non-college graduates (88% versus 64%). One might
therefore expect larger gains in coverage among males and non-college graduates. On the other
hand, young adults of high socioeconomic status may be more likely to have parents with
employer-provided coverage, so the gains in coverage could potentially be larger for college
graduates.
A second possible source of heterogeneity is that, even if the gains in health insurance are
the same among all groups, different groups could respond differently to receiving coverage. For
instance, Grossman (1972) argues that education enables individuals to become more efficient
producers of health. More education may therefore better equip individuals to make the most out
of the newly-acquired insurance (e.g. more easily find providers who accept the insurance, ask
better questions at doctors appointments, or better follow medical advice). Alternatively, the
price elasticity of medical care could be strongest among low-income individuals, in which case
the effects of obtaining insurance on health care utilization and health could be largest for noncollege graduates. The price elasticity of medical care could also differ by sex. For instance,
evidence suggests that females are more risk averse than males (e.g. Jiankoplos and Bernasek,
1998). One might therefore expect females to be more likely to obtain medical care regardless of
its price, whereas males might only utilize care if the cost is minimal; i.e. males might have
stronger price elasticities. Indeed, in our pre-treatment data uninsured females had higher rates of
primary care doctor access, flu vaccination, and well-patient checkups than uninsured males.
Third, as discussed at the beginning of Section IV, the dependent coverage provision
could affect health-related outcomes through mechanisms besides the extensive margin of

25

insurance coverage particularly the intensive margin of coverage and there could be
heterogeneous effects along these dimensions. For instance, suppose part of the reason females
and college graduates had lower pre-ACA uninsured rates was because they were more likely to
privately purchase a bare-bones, catastrophic plan if they did not have access to employerprovided coverage. In that case, the ACA dependent coverage provision may lead to larger gains
along the intensive margin of coverage for women and college graduates, leading to larger
improvements in health-related outcomes among these groups.
The first two columns of Table 1.6 report the results for females and males. Males
experienced a 2.9 percentage point larger gain in health insurance coverage than females, and the
difference is significant at the 1% level. Moreover, only males experienced statistically
significant favorable effects on any outcomes besides insurance coverage. Specifically, males
rates of primary care doctor access, having any exercise, reporting very good/excellent health,
and reporting excellent health increased substantially by 4.6, 1.9, 2.9, and 3.1 percentage points,
respectively. These effects are all significantly different from zero, and three of the four (all but
very good/excellent health) are also statistically different from the corresponding effects on
females. The only statistically significant result for females (besides insurance coverage) is an
adverse effect on days with health-related limitations. In sum, the results suggest that males
experienced larger improvements in health-related outcomes from the ACA dependent coverage
provision than females, and that there appear to be multiple reasons for this heterogeneity. Gains
in insurance coverage were larger for males, consistent with them having a higher pre-ACA
uninsured rate. Responses to obtaining insurance coverage also appear to have been stronger for
males, perhaps indicating a larger price elasticity of demand for medical care.

26

The last two columns of Table 1.6 report the results stratifying by college degree
attainment. Both groups experienced similar gains in insurance coverage as a result of the ACA
dependent coverage provision. However, statistically significant improvements in outcomes
besides health insurance are only observed for college graduates. The mandate led to large and
significant gains for college graduates in the following outcomes: primary care doctor access (5.1
percentage points), cost being a barrier to care (reduction of 3.4 percentage points), BMI
(reduction of 0.25 units), obesity (reduction of 1.7 percentage points), and excellent self-reported
health (increase of 3.7 percentage points). Besides insurance, the only significant effects for noncollege graduates are unfavorable: a 2.2 percentage point reduction in flu vaccinations and a 1.6
percentage point increase in risky drinking. In short, college graduates experienced greater
improvements in health-related outcomes than non-college graduates, and this appears to be due
to heterogeneous effects of coverage rather than heterogeneous effects on coverage. This is
consistent with a Grossman-style story in which education enables individuals to better take
advantage of their health care opportunities. However, the results could also be partly
attributable to greater gains along the intensive margin of coverage for college graduates, which
we cannot measure in our data. Regardless of the reason, these results suggest that the mandate
increases SES-based disparities in health.
VII. Discussion
The first major insurance expansion under the ACA a provision requiring insurers to
allow young adults to remain on their parents health insurance until turning 26 was
implemented in September 2010. This paper uses data from the BRFSS to examine the effects of
this mandate on various outcomes related to health care access, preventive care utilization, risky
health behaviors, and self-assessed health. We implement a difference-in-differences model with

27

individuals slightly below the mandates age cutoff (ages 23-25) as the treatment group and those
slightly above the cutoff (ages 27-29) as the control group.
We first estimate average effects for the entire sample. The results suggest that the ACA
dependent coverage provision increased health care access but not utilization of preventive care,
had mixed effects on risky health behaviors, and improved health at the high end of the
distribution. Specifically, we observe significant and robust favorable effects on health insurance,
access to a primary care doctor, probability of having excellent self-assessed health, and BMI.
However, we also find an adverse effect on risky drinking consistent with ex ante moral hazard
and no clear effects on the other outcomes. We then validate our model through a series of
placebo tests and show that our classifications of treatment and control groups perform better in
these tests that the wider age bandwidths common in the literature. Finally, we conduct
subsample analyses, finding particularly striking improvements in outcomes for men and college
graduates. Men had larger gains in health insurance coverage than women, and only men
experienced statistically significant gains in any outcomes beyond health insurance specifically
primary care access, exercise, and overall self-assessed health. Insurance expansions were
similar for college graduates and non-college graduates, but only college graduates experienced
significant gains in any other outcomes: primary care access, cost being a barrier to care, BMI,
obesity, and overall self-assessed health.
The ACA dependent coverage mandate provides a unique opportunity to study a health
insurance intervention specific to young adults as opposed to seniors (Medicare), the poor
(Medicaid), or the uninsured population at large (the Massachusetts reform). In general, our
results suggest that health insurance affects health-related outcomes of young adults more
modestly than prior studies have observed for these other populations. First, we find no evidence

28

of increased preventive care utilization, in contrast to prior results from both Medicaid
(Finkelstein et al., 2012) and the Massachusetts reform (Kolstad and Kowalski, 2012). Second,
we only find statistically significant improvements in overall self-assessed health at the top of the
distribution, as reporting of excellent health increases but there is no clear evidence of an effect
on reporting very good or excellent health. We do not observe any gains in the variables
reflecting more severe health conditions: days not in good physical health, days not in good
mental health, and days with functional limitations. This contrasts the clear gains in these same
outcomes observed for both Medicaid (Finkelstein et al., 2012) and the Massachusetts reform
(Van der Wees et al., 2013; Courtemanche and Zapata, 2014). Interestingly, Chua and Sommers
(2014) find that the ACA dependent coverage provision increased the probabilities of selfreporting excellent physical and mental health. Combining their results with ours suggests that
physical and mental health did improve, but only at the high end of the distribution.
While our results suggest that health insurance expansions for young adults are less
impactful than those for other age groups, it is still important to emphasize that we do observe
some improvements in important outcomes, including health care access, excellent self-assessed
health, and BMI. One might have initially worried that a coverage expansion for young adults
would not lead to any health improvements given the generally good baseline health of this age
group.
An important contribution of our paper is that we provide, to our knowledge, the first
empirical investigation of ex ante moral hazard that focuses specifically on young adults. We
find evidence consistent with ex ante moral hazard in only one domain: risky drinking (binge
drinking or excessive number of drinks per month). In contrast, we find evidence that the

29

dependent coverage improved weight-related behaviors while not affecting smoking and
pregnancies. Our results therefore suggest that ex ante moral hazard is domain-specific.
Another interesting result is that, since the improvement in health is concentrated among
college graduates, the ACA dependent coverage provision appears to increase SES-based
disparities in health. This is contrary to the usual impacts of public policies to expand health
insurance. Medicaid has been shown to improve at least some health outcomes (Currie and
Gruber 1996a and 1996b; Finkelstein et al., 2012; Sommers et al., 2012), implying reduced
income-based disparities in health. The Massachusetts reform also appears to have reduced
income-based disparities, as Courtemanche and Zapata (2014) found the largest gains in selfassessed health among low-income individuals.
Several caveats to our analyses provide directions for future research. First, since we
study eighteen different dependent variables, we might expect one or two results to emerge as
significant at conventional levels simply by chance. We did not employ multiple hypothesis test
adjustments in this paper because, even though such adjustments control the Type I error rate
(probability of falsely rejecting any null hypotheses), they do so at the cost of substantially
increasing the Type II error rate (probability of failing to reject false null hypotheses).21 However,
future research should revisit our questions using different data to see if any of our findings
could be attributable to chance rather than genuine causal effects of the mandate.

21

For instance, the simple Bonferroni correction involves multiplying all p-values by the number of hypotheses
being tested, which is eighteen in our case. This would make it virtually impossible to reject any null hypothesis in
regressions that already demand quite a bit of the data by including fixed effects and clustering at an aggregated
level. It is not clear to us that it would be preferable to, for example, fail to reject five false null hypotheses for the
sake of not rejecting one true null hypothesis. This seems especially true in cases such as ours, where null results are
an important part of the story. Moreover, we view our analyses as testing for eighteen distinct effects, some of which
are more plausible theoretically than others, as opposed to testing for one effect that may manifest itself through
eighteen different measures. It is not clear why, for instance, we should inflate the p-values in the health insurance
regressions merely because we also study smoking, pregnancies, etc.

30

Next, we focus on estimating the ACA dependent coverage provisions effects on 23-25
year olds, ignoring possible effects on 19-22 year olds because of the greater difficulty in finding
a suitable control group and the weaker ex ante expectations of significant effects. Further
understanding whether benefits accrue to young adults besides 23-25 year olds is obviously
important in order to fully evaluate the policy.
Further research is also necessary to understand the mechanisms through which the
mandate improves health. Increased health care utilization is an obvious possibility, but early
evidence on the ACA provisions impact on health care consumption is mixed. Akosa Antwi et
al. (2014) report a rise in hospitalizations using administrative data, but Chua and Sommers
(2014) find no evidence of changes in survey-based measures of hospital care, primary care, or
prescription drug utilization, while we find no significant increases in preventive care. Another
possible explanation is that self-assessments of health improve due to a warm glow from the
peace of mind of having insurance. Finkelstein et al. (2012) proposed this as an explanation for
their finding from the Oregon Medicaid experiment that most of the gains in self-assessed health
appeared to occur before changes in utilization.
Finally, and critically, our results should not be interpreted as providing a full accounting
of the benefits of expanding insurance coverage among young adults. The primary purpose of
insurance is to protect individuals from financial risk, and gains along this dimension may be
especially substantial for young adults given their relatively low income and wealth levels.
Moreover, expanding coverage among young adults is an important component of the overall
strategy behind the ACA since it is necessary to offset the additional costs of insuring older and
sicker individuals under community rating. In other words, the costs and benefits of the different

31

components of the ACA need to be evaluated together, as the different pieces of the reform are
designed to work synergistically.

32

Table 1.1 Sample Sizes for Different Outcomes

Outcome Variable
Health care access
Any health insurance coverage
Any primary care doctor
Cost prevented care in past year
Preventive care utilization
Flu vaccination in past year
Well-patient checkup in past year
Pap test in past year (women only)+
Risky health behaviors
Currently smokes cigarettes
Alcoholic drinks in past 30 days
Risky drinker in past 30 days
Body mass index
Obese
Any exercise in past 30 days
Pregnancy (unmarried women only)
Self-assessed health
Overall health very good or excellent
Overall health excellent
Days of last 30 not in good mental health
Days of last 30 not in good physical health
Days of last 30 with health-related limitations
+

Total

Treatment
(23-25)

Control
(27-29)

126,702
118,392
107,831

53,057
49,520
45,041

73,645
68,872
62,790

118,394
107,931
26,919

49,502
45,085
10,799

68,892
62,846
16,120

125,616
120,958
120,037
120,373
120,373
122,720
39,499

52,607
50,521
50,110
50,529
50,529
51,337
19,610

73,009
70,437
69,927
69,844
69,844
71,383
19,889

126,662
126,662
124,773
124,861
125,365

53,102
53,102
52,386
52,387
52,615

73,560
73,560
72,387
72,474
72,750

The pap test variable is only available in even-numbered years, reducing the sample size for that outcome.

33

Table 1.2 Pre-Treatment Means and Standard Deviations for Control Variables
Control Variable
Treatment (Ages 23-25) Control (Ages 27-29)
Age dummies (age=23 is omitted)
Age=24
0.349 (0.477)
-Age=25
0.322 (0.467)
-Age=27
-0.310 (0.462)
Age=28
-0.343 (0.475)
Age=29
-0.347 (0.476)
Female
0.505 (0.500)
0.508 (0.500)
Race/ethnicity dummies (non-Hispanic white is omitted)
Non-Hispanic black
0.112 (0.316)
0.116 (0.320)
Hispanic
0.224 (0.417)
0.209 (0.407)
Other than black, Hispanic, or white
0.087 (0.282)
0.077 (0.266)
Currently married
0.305 (0.460)
0.564 (0.496)
Education dummies (less than high school degree is omitted)
High school degree but no further
0.283 (0.450)
0.257 (0.437)
Some college but no four-year degree
0.299 (0.458)
0.271 (0.444)
College graduate
0.303 (0.459)
0.364 (0.481)
Household income dummies (less than $10,000 is omitted)
Between $10,000 and $15,000
0.068 (0.252)
0.049 (0.216)
Between $15,000 and $20,000
0.102 (0.303)
0.077 (0.267)
Between $20,000 and $25,000
0.116 (0.321)
0.097 (0.296)
Between $25,000 and $35,000
0.144 (0.351)
0.129 (0.335)
Between $35,000 and $50,000
0.166 (0.372)
0.165 (0.371)
Between $50,000 and $75,000
0.143 (0.350)
0.187 (0.390)
$75,000 and over
0.186 (0.389)
0.240 (0.427)
Number of children in household dummies (0 is omitted)
One child
0.230 (0.421)
0.235 (0.424)
Two children
0.159 (0.366)
0.233 (0.423)
Three children
0.055 (0.229)
0.110 (0.313)
Four children
0.018 (0.133)
0.038 (0.192)
Five or more children
0.008 (0.090)
0.016 (0.124)
+
0.678 (0.467)+
Cell phone only
0.703 (0.457)
Student
0.109 (0.312)
0.054 (0.226)
Unemployed
0.111 (0.314)
0.093 (0.290)
State unemployment rate
7.032 (2.615)
7.186 (2.666)
Pre-ACA state mandate
0.220 (0.415)
0.033 (0.179)
Notes: BRFSS sampling weights are used. Means are reported, with standard deviations in parentheses. + indicates
the summary statistics are from 2011-2013, since the variable is 0 for all respondents in all prior years.

34

Table 1.3 Means and Standard Deviations for Outcome Variables

Outcome Variable
Health care access
Any health insurance coverage

Pre-Treatment Period
Treatment (Ages Control (Ages 2723-25)
29)

Post-Treatment Period
Treatment (Ages
23-25)

Control (Ages 2729)

Difference-inDifferences

0.680 (0.466)

0.753 (0.431)

0.709 (0.454)

0.708 (0.455)

0.073 (0.018)***

0.564 (0.496)
0.241 (0.427)

0.641 (0.480)
0.216 (0.411)

0.519 (0.500)
0.240 (0.427)

0.558 (0.497)
0.235 (0.424)

0.038 (0.010)**
-0.020 (0.014)

0.225 (0.418)
0.521 (0.500)
0.693 (0.461)

0.246 (0.431)
0.545 (0.498)
0.724 (0.447)

0.239 (0.426)
0.524 (0.499)
0.614 (0.487)

0.265 (0.441)
0.529 (0.499)
0.647 (0.478)

-0.006 (0.009)
0.019 (0.011)
-0.002 (0.013)

Risky health behaviors

Currently smokes cigarettes
Alcoholic drinks in past 30 days

0.260 (0.432)
17.359 (43.926)

0.249 (0.432)
13.883 (34.703)

0.257 (0.437)
19.481 (43.947)

0.254 (0.435)
16.841 (40.916)

-0.009 (0.012)
-0.836 (0.889)

Risky drinker in past 30 days

Body mass index

0.775 (0.418)
26.404 (5.807)

0.807 (0.394)
27.253 (6.031)

0.749 (0.434)
26.167 (6.019)

0.769 (0.422)
27.192 (6.142)

0.013 (0.005)*
-0.177 (0.050)**

Obese
Any exercise in past 30 days

0.222 (0.415)
0.810 (0.392)

0.262 (0.440)
0.799 (0.401)

0.197 (0.398)
0.819 (0.385)

0.252 (0.434)
0.799 (0.401)

-0.014 (0.003)***
0.009 (0.005)

Pregnancy

0.048 (0.215)

0.043 (0.203)

0.044 (0.205)

0.040 (0.195)

-0.001 (0.004)

Overall health very good/excellent

Overall health excellent

0.607 (0.488)
0.255 (0.436)

0.610 (0.488)
0.257 (0.437)

0.608 (0.488)
0.250 (0.433)

0.589 (0.492)
0.236 (0.425)

0.022 (0.011)
0.017 (0.003)***

Days not in good mental health

Days not in good physical health

4.050 (7.638)
2.240 (5.526)

3.844 (7.680)
2.303 (5.815)

4.410 (8.067)
2.446 (5.999)

4.165 (8.063)
2.484 (6.170)

0.040 (0.162)
0.025 (0.053)

Days with health-related limitations

1.589 (4.757)

1.664 (5.177)

1.727 (5.131)

1.739 (5.332)

0.063 (0.104)

Any primary care doctor

Cost prevented care in past year
Preventive care utilization
Flu vaccination in past year
Well-patient checkup in past year
Pap test in past year

Self-assessed health

Notes: Standard errors, heteroskedasticity-robust and clustered by age, are in parentheses. BRFSS sampling weights are used. Means are reported, with standard
deviations in parentheses. *** indicates the difference-in-difference is significant at the 1% level; ** 5% level; * 10% level.

35

Table 1.4 Difference-in-Difference Regression Estimates of Effects of ACA Dependent Coverage Mandate
Baseline Model

Demographic
Controls Only

Any health insurance

0.061 (0.017)** [0.130]

0.067 (0.018)**

0.059 (0.013)*** 0.055 (0.012)*** 0.064 (0.016)***

0.061 (0.015)***

Any primary doctor

0.032 (0.010)** [0.065]

0.034 (0.010)**

0.020 (0.006)**

0.021 (0.006)**

0.033 (0.009)**

0.029 (0.011)**

Cost prevented care

-0.019 (0.014) [-0.044]

-0.019 (0.014)

-0.022 (0.015)

-0.023 (0.015)

-0.020 (0.015)

-0.016 (0.011)

-0.014 (0.007) [-0.033]

-0.011 (0.008)

-0.017 (0.009)

-0.018 (0.008)*

-0.014 (0.008)

-0.020 (0.006)**

0.013 (0.011) [0.026]

0.015 (0.010)

0.011 (0.010)

0.011 (0.010)

0.017 (0.010)

0.011 (0.006)

-0.004 (0.015) [-0.009]

-0.003 (0.014)

-0.019 (0.010)

-0.025 (0.015)**

-0.015 (0.015)

0.002 (0.008)

Currently smokes

0.003 (0.007) [0.007]

-0.006 (0.010)

-0.001 (0.004)

-0.008 (0.005)

-0.001 (0.006)

0.005 (0.007)

Drinks per month

0.120 (0.906) [0.003]

-0.468 (0.887)

-0.429 (0.604)

-0.597 (0.590)

0.083 (0.840)

0.011 (0.929)

0.011 (0.003)** [0.026]

0.008 (0.004)*

0.009 (0.003)**

0.009 (0.007)**

0.014 (0.003)***

0.009 (0.003)**

-0.124 (0.062)

-0.169 (0.061)**

-0.173 (0.074)*

-0.118 (0.033)***

Outcome Variable

Start in 2004

Start in 2001

Drop 3/10-12/10

Collapsed Data

Health care access

Preventive care utilization

Flu vaccination
Well-patient checkup
Pap test
Risky health behaviors

Risky drinker
Body mass index
Obese

-0.098 (0.029)** [-0.017] -0.175 (0.045)**

-0.009 (0.008) [-0.022]

-0.014 (0.005)**

-0.010 (0.007)

-0.011 (0.008)

-0.013 (0.006)*

-0.010 (0.004)**

0.003 (0.004) [0.008]

0.008 (0.007)

0.005 (0.003)

0.004 (0.004)

0.001 (0.005)

0.007 (0.003)**

-0.003 (0.005) [-0.014]

-0.002 (0.005)

-0.004 (0.004)

-0.002 (0.004)

-0.003 (0.005)

-0.002 (0.004)

0.015 (0.011) [0.031]

0.018 (0.010)

0.016 (0.011)

0.015 (0.008)

0.011 (0.010)

0.014 (0.009)

0.014 (0.005)** [0.032]

0.014 (0.003)***

0.013 (0.004)**

0.014 (0.005)**

0.014 (0.006)*

0.015 (0.004)***

Days not good mental

0.081 (0.158) [0.010]

0.064 (0.144)

0.050 (0.144)

0.036 (0.116)

0.156 (0.156)

0.084 (0.127)

Days not good phys.

0.059 (0.068) [0.011]

0.122 (0.099) [0.025]

0.045 (0.046)
0.102 (0.093)

-0.022 (0.079)
0.073 (0.109)

-0.014 (0.076)
0.065 (0.094)

0.075 (0.028)**
0.201 (0.107)

0.028 (0.063)
0.101 (0.086)

Any exercise
Pregnancy
Self-assessed health
Very good/exc. Health
Excellent health

Days health limitations

Notes: *** indicates significant at the 1% level; ** 5% level; * 10% level. Standard errors, heteroskedasticity-robust and clustered by age, are in parentheses. All
regressions include the controls plus age, state, and time fixed effects. BRFSS sampling weights are used. For the baseline regression, effect sizes in standard
deviations of the dependent variable (for the treatment group in the pre-treatment period) are in brackets.

36

Table 1.5 Placebo Regressions

Outcome Variable
Health care access
Any health insurance coverage
Any primary care doctor
Cost prevented care in past year
Preventive care utilization
Flu vaccination in past year
Well-patient checkup in past year
Pap test in past year
Risky health behaviors
Currently smokes cigarettes
Alcoholic drinks in past 30 days
Risky drinker in past 30 days
Body mass index
Obese
Any exercise in past 30 days
Pregnancy
Self-assessed health
Overall health very good/excellent
Overall health excellent
Days not in good mental health
Days not in good physical health
Days with health-related limitations

2003-2009
2002-2008
2001-2007
Treatment 10/06 Treatment 10/05 Treatment 10/04
-0.002 (0.005)
-0.008 (0.014)
-0.007 (0.011)

0.002 (0.008)
0.002 (0.007)
-0.017 (0.013)

-0.009 (0.007)
0.019 (0.011)
-0.013 (0.007)

-0.013 (0.016)
0.002 (0.014)
-0.012 (0.014)

-0.007 (0.008)
-0.008 (0.014)
-0.024 (0.022)

-0.001 (0.007)
--0.027 (0.025)

-0.019 (0.007)**
-1.648 (1.035)
-0.001 (0.006)
-0.001 (0.146)
0.002 (0.008)
0.008 (0.008)
0.011 (0.010)

-0.006 (0.009)
-0.659 (0.584)
-0.014 (0.007)*
-0.023 (0.196)
0.005 (0.011)
0.008 (0.005)
0.005 (0.012)

-0.007 (0.008)
-1.146 (0.788)
-0.009 (0.011)
-0.082 (0.145)
0.0005 (0.009)
-0.004 (0.005)
-0.002 (0.007)

0.011 (0.003)**
0.004 (0.005)
-0.064 (0.174)
-0.041 (0.109)
-0.039 (0.084)

0.002 (0.009)
0.009 (0.005)
-0.005 (0.232)
0.034 (0.121)
-0.043 (0.051)

0.004 (0.009)
0.006 (0.008)
-0.054 (0.143)
0.165 (0.107)
0.017 (0.060)

Notes: *** indicates significant at the 1% level; ** 5% level; * 10% level. Standard errors, heteroskedasticity-robust
and clustered by age, are in parentheses. All regressions include the controls plus age, state, and time fixed effects.
BRFSS sampling weights are used.

37

Table 1.6 Heterogeneity by Sex and Education

Sex
Outcome Variable

Female

Male

Education
Not College
College Graduate
Graduate

Health care access

Any health insurance coverage
0.045 (0.017)** 0.074 (0.016)***+++ 0.067 (0.019)** 0.061 (0.013)***
Any primary care doctor
0.016 (0.009)
0.046 (0.012)**+
0.025 (0.012)
0.051 (0.006)***++
Cost prevented care in past year
-0.019 (0.021)
-0.016 (0.013)
-0.014 (0.017)
-0.034 (0.009)**
Preventive care utilization
Flu vaccination in past year
-0.020 (0.012)
-0.012 (0.010)
-0.022 (0.008)**
0.003 (0.009)
Well-patient checkup in past year
0.013 (0.014)
0.013 (0.016)
0.006 (0.019)
0.035 (0.016)
Pap test in past year
-0.004 (0.015)
--0.007 (0.021)
0.008 (0.028)
Risky health behaviors
Currently smokes cigarettes
0.011 (0.011)
-0.004 (0.016)
0.001 (0.008)
0.002 (0.006)
Alcoholic drinks in past 30 days
-0.117 (0.441)
0.359 (1.559)
-0.068 (1.171)
0.398 (0.985)
Risky drinker
0.009 (0.012)
0.015 (0.014)
0.016 (0.004)***
-0.007 (0.007)
Body mass index
-0.133 (0.153)
0.018 (0.160)
0.001 (0.050)
-0.254 (0.096)**
Obese
-0.010 (0.010)
-0.005 (0.012)
-0.004 (0.009)
-0.017 (0.004)***
++
Any exercise in past 30 days
-0.010 (0.007) 0.019 (0.004)***
0.001 (0.006)
0.010 (0.005)
Pregnancy
-0.003 (0.005)
--0.005 (0.006)
-0.001 (0.006)
Self-assessed health
Overall health very good or excellent
0.001 (0.022)
0.029 (0.009)**
0.007 (0.009)
0.029 (0.017)
++
Overall health excellent
-0.003 (0.009) 0.031 (0.005)***
0.002 (0.006)
0.037 (0.012)**
Days of last 30 not in good mental health
0.100 (0.196)
0.083 (0.160)
0.259 (0.154)
-0.323 (0.193)+++
Days of last 30 not in good physical health
0.109 (0.081)
-0.011 (0.167)
0.211 (0.145)
-0.262 (0.166)
Days of last 30 with health-related limitations 0.347 (0.110)**
-0.102 (0.206)
0.265 (0.161)
-0.149 (0.096)
Notes: +++ difference between effects on subgroups is significant at the 1% level; ++ 5% level; + 10% level. See other notes for Table 1.5.

38

Table 1.7 Full Regression Output for Selected Dependent Variables

Control Variable
Insurance
Smoker
Treated*Post
0.061 (0.017)*
0.003 (0.007)
Age=24
-0.004 (0.002)*
0.007 (0.001)**
Age=25
-0.007 (0.003)
0.022 (0.001)**
Age=27
0.012 (0.010)
0.030 (0.002)**
Age=28
0.016 (0.011)
0.039 (0.002)**
Age=29
0.029 (0.011)*
0.032 (0.002)**
Female
0.070 (0.006)**
-0.056 (0.005)**
Non-Hispanic black
-0.012 (0.010)
-0.126 (0.009)**
Hispanic
-0.117 (0.007)** -0.183 (0.004)**
Other than black, Hispanic, or white
-0.011 (0.010)
-0.025 (0.013)
Currently married
0.069 (0.008)**
-0.110 (0.008)**
High school degree but no further
0.110 (0.012)**
-0.091 (0.017)**
Some college but no 4-year degree
0.171 (0.017)**
-0.161 (0.014)**
College graduate
0.251 (0.019)**
-0.310 (0.020)**
Between $10,000 and $15,000
-0.048 (0.014)*
0.007 (0.006)
Between $15,000 and $20,000
-0.070 (0.018)*
0.021 (0.011)
Between $20,000 and $25,000
-0.032 (0.009)*
0.007 (0.009)
Between $25,000 and $35,000
0.051 (0.010)**
-0.020 (0.013)
Between $35,000 and $50,000
0.120 (0.009)**
-0.036 (0.014)
Between $50,000 and $75,000
0.169 (0.011)**
-0.063 (0.016)**
$75,000 and over
0.179 (0.012)**
-0.057 (0.015)*
One child in household
0.021 (0.007)*
0.035 (0.009)**
Two children in household
0.031 (0.005)**
0.044 (0.012)*
Three children in household
0.020 (0.008)
0.055 (0.010)**
Four children in household
0.017 (0.021)
0.071 (0.023)*
Five or more children in household
0.065 (0.023)*
0.071 (0.018)*
Cell phone only
-0.013 (0.006)
0.007 (0.006)
Student
-0.006 (0.016)
-0.035 (0.008)**
Unemployed
-0.164 (0.017)**
0.100 (0.010)**
State unemployment rate
0.004 (0.003)
-0.009 (0.004)*
Pre-ACA state mandate
0.017 (0.010)
0.001 (0.010)

Excellent Health
0.014 (0.005)*
-0.002 (0.001)*
-0.004 (0.001)**
-0.016 (0.002)**
-0.026 (0.002)**
-0.018 (0.003)**
-0.020 (0.008)
0.006 (0.007)
-0.019 (0.005)**
-0.010 (0.004)*
0.027 (0.005)**
0.039 (0.009)**
0.051 (0.008)**
0.111 (0.009)**
-0.008 (0.018)
-0.007 (0.013)
0.0001 (0.009)
0.026 (0.013)
0.044 (0.010)**
0.058 (0.011)**
0.108 (0.012)**
-0.012 (0.003)**
-0.010 (0.008)
-0.025 (0.011)
-0.038 (0.016)
-0.007 (0.012)
0.011 (0.007)
0.013 (0.009)
-0.027 (0.006)**
0.002 (0.002)
-0.015 (0.004)*

Notes: ** indicates significant at the 1% level; * 5% level. Standard errors, heteroskedasticity-robust and clustered
by age, are in parentheses. All regressions also include the age, state, and time fixed effects. BRFSS sampling
weights are used. Separate variables for treated and post are not included because they are subsumed by the age
and time fixed effects.

39

Figure 1.1 -- Trends in Access to Care and Preventive Care Variables by Age Group

.25

.5

.1

.55

.15

.6

.2

.65

.8
.75
.7
.65
2001 2003 2005 2007 2009 2011 2013
Fit line

2001 2003 2005 2007 2009 2011 2013

Age27~29

Age23~25

95% CIs

Fit line

2001 2003 2005 2007 2009 2011 2013

Age27~29

Age23~25

95% CIs

Fit line

Patient Checkups

95% CIs

Pap Tests

2001 2003 2005 2007 2009 2011 2013

Age23~25
Fit line

Age27~29
95% CIs

.75
.65
.6

.5

.15

.52

.2

.7

.54

.25

.56

.3

Flu Shots

Age27~29

.8

Age23~25

Cost Barrier to Care

.3

Personal Doctor Visits

.7

Health Insurance Coverage

2001 2003 2005 2007 2009 2011 2013

Age23~25
Fit line

40

Age27~29
95% CIs

2001 2003 2005 2007 2009 2011 2013

Age23~25
Fit line

Age27~29
95% CIs

Figure 1.2 -- Trends in Health Behavior Variables by Age Group

Alcoholic Drinks per Month

.85
.75

15

.7

10
2001 2003 2005 2007 2009 2011 2013

Risky Drinkers

.8

20

25

.22 .24 .26 .28 .3 .32

Smokers

2001 2003 2005 2007 2009 2011 2013

2001 2003 2005 2007 2009 2011 2013

23~25

27~29

23~25

27~29

23~25

27~29

Fit line

95% CIs

Fit line

95% CIs

Fit line

95% CIs

Any Exercise
.78 .8 .82 .84 .86

Obesity

25

.15

.2

26

27

.25

.3

28

Body Mass Index

2001 2003 2005 2007 2009 2011 2013

2001 2003 2005 2007 2009 2011 2013

2001 2003 2005 2007 2009 2011 2013

23~25

27~29

23~25

27~29

23~25

27~29

Fit line

95% CIs

Fit line

95% CIs

Fit line

95% CIs

.02 .03 .04 .05 .06 .07

Pregnancy

2001 2003 2005 2007 2009 2011 2013

23~25

27~29

Fit line

95% CIs

41

Figure 1.3 -- Trends in Self-Assessed Health Variables by Age Group

Overall Health - Excellent

4.5

.28

3.5

.26
.24
Age23~25
Fit line

2001 2003 2005 2007 2009 2011 2013

Age27~29

Age23~25

95% CIs

Fit line

Days Not in Good Physical Health

Age27~29
95% CIs

Days with Functional Limitations

1 1.2 1.4 1.6 1.8

1.8 2 2.2 2.4 2.6 2.8

.22
2001 2003 2005 2007 2009 2011 2013

2001 2003 2005 2007 2009 2011 2013

Age23~25
Fit line

Age27~29
95% CIs

Days Not in Good Mental Health

5

.3

.58 .6 .62 .64 .66 .68

Overall Health - Very Good & Excellent

2001 2003 2005 2007 2009 2011 2013

Age23~25
Fit line

42

Age27~29
95% CIs

2001 2003 2005 2007 2009 2011 2013

Age23~25
Fit line

Age27~29
95% CIs

CHAPTER II
Health Insurance and Young Adults Avoidable Hospitalizations
I. Introduction
On March 23, 2010, President Obama signed the Patient Protection and Affordable Care
Act (ACA) into law. 22 One of the first implemented provisions of the ACA was targeted at
young adults, who often face the risk of losing their health insurance coverage as early as age 19.
Prior to the ACA, insurance companies typically removed enrolled children from their parents
plans at age 19 for non-students and 23 for full-time students (Anderson et al., 2012 and 2014).23
Under the new law, starting in September 2010, young adults are allowed to stay on their parents
plan until they turn 26 years old, with the same benefits.24 By allowing young adults to maintain
coverage under their parents health plan, the law makes it easier and more affordable for them
to get health care.
Historically, the rate of insurance coverage for young Americans decreased at age 19, as
these young adults may have lost their health insurance due to being ineligible to maintain
coverage under their parents plan or because of their employment status (unemployed, part-time
employment, entry-level employment or small business employment without employersponsored coverage). For these reasons, young adults typically have the lowest rate of insurance
coverage in comparison with other age groups. To be more specific, the rate of insurance
coverage for young adults in the age group of 19-25 was only 68.6 percent in 2009, while the
national rate was 83.9 percent (DeNavas-Walt et al., 2010).
22

For more information one can visit the following websites:

http://www.whitehouse.gov/healthreform/healthcare-overview
http://www.hhs.gov/healthcare/rights/law/index.html
23
There was a great deal of prior state-to-state variation in dependent coverage rules, including differences in age
limits and marital status requirements.
24
For more on this policy see: http://www.hhs.gov/healthcare/rights/youngadults/index.html;
http://www.cms.gov/CCIIO/Resources/Files/adult_child_fact_sheet.html.

43

Contrary to the idea that young people do not need health insurance, one out of six
young adults experiences a chronic illness like cancer, asthma or appendicitis (Centers for
Disease Control and Prevention, 2009). Also, young adults often partake in behaviors such as
overeating, sedentary lifestyles, smoking, excessive drinking, and unprotected sex that pose
long-term risks. Additionally, compared to insured young adults, uninsured peers are two-to-four
times more likely to delay healthcare due to costs (Cantor, 2010). Moreover, young adults are at
risk for their health as well as their finances: nearly half of uninsured young adults report
problems associated with paying medical bills (Collins, 2012). Lacking health insurance as a
young adult tends to cause health and economic problems in later adulthood (Merluzzi, 1999;
Callahan, 2005; Nicholson, 2009).
A recent literature has developed showing that the ACA expansion of dependent
coverage increased the rate of insurance coverage among the targeted group of young adults
(Cantor et al., 2012; Sommers and Kronick, 2012; Sommers et al., 2013; Akosa Antwi et al.,
2013 and 2015; Chua and Sommers, 2014; Barbaresco et al., 2015). However, there is little, if
any, evidence on the effect of this aspect of the ACA on the quality of care received by young
adults. The purpose of this paper is to evaluate the impact of the ACA expansion of dependent
coverage on primary care quality by examining changes over time in the probability of having an
avoidable hospitalization among the targeted group of young adults as compared to young adults
just outside this age range.
As in the Kolstad and Kowalski (2012) (hereafter as KK) study of the Massachusetts
health care reform, I analyze the universe of hospital discharges from a nationally-representative
sample of roughly 20 percent of all hospitals in the United States that is compiled by the Agency
for Healthcare Research and Quality (AHRQ) Healthcare Cost and Utilization Project (HCUP).

44

This sample is known as the National Inpatient Sample (NIS). In addition, I also follow KK and
use the AHRQ-provided methodology for identifying avoidable hospitalizations in the data.
While conventional wisdom suggests that an increase in insurance coverage would result in more
preventive care in an outpatient setting and thus fewer avoidable hospitalizations, I develop a
conceptual model in which the impact of the expansion of dependent coverage on the probability
of having an avoidable hospitalization is ambiguous and use that to motivate my empirical work.
In contrast to KKs findings for the Massachusetts reform, my primary results suggest that the
ACA expansion of dependent coverage for young adults aged 23-25 leads to an increase in
overall avoidable hospitalizations, which is driven by a large increase in chronic avoidable
hospitalizations. The effects are stronger for female, whites, and the middle income quartiles of
patients zip code.
Some would interpret this result as implying the ACA led a reduction in primary care
quality. It may instead suggest a tradeoff between two forces likely to increase avoidable
hospitalization rates, the moral hazard aspect of expanding insurance coverage as well as
improved access to hospitals, and the efficiency effect associated with increasing access to
primary care relative to hospital and emergency room care, which should reduce avoidable
hospitalization rates. The size of this tradeoff is likely different for the young adults targeted by
this reform as compared to older adults, children, or the elderly.
The rest of this paper is organized as follows: Section II provides an overview of the
relevant literature, section III describes the conceptual relationship between the ACA expansion
of dependent insurance coverage for young adults and the number of avoidable hospitalizations,
section IV describes my methodology, and section V describes the data. Section VI presents my

45

results and section VII provides a discussion of these results. Conclusions are given in section
VIII.
II. Literature Review
In this section I review the literature on previous state policy as well as the ACA mandate
with respect to dependent coverage and related types of coverage expansions. I focus on the
literature dealing with the impacts of expansions of coverage on general health care utilization,
risky behaviors, and health outcomes, as well as avoidable hospitalizations.
A. Dependent Coverage Policies and Insurance Coverage
Prior to the ACA, about two thirds of states implemented state-level policies allowing for
some type of dependent coverage expansion. However, researchers found small or even no net
impact of these policies on the number of uninsured young people (Levine et al., 2011; Monheit
et al., 2011; Blum et al., 2012). This was due in part to the scope of these reforms being limited
by the state definitions of a dependent, which could include restrictions related to student status,
marital status, co-residence with parents and tax dependent status. Additionally, all state laws
excluded self-funded benefit programs, which meant that they did not apply to around half of
employer-provided plans. In addition, increases in dependent coverage could have been offset by
reductions in other types of coverage.
In contrast, the ACA dependent coverage expansion aimed to improve net coverage
among young adults by relaxing the eligibility requirements and extending the same
requirements to employers who have self-insured plans. Recent studies have shown that the ACA
dependent coverage expansion has significantly increased health insurance coverage levels for
young adults across all racial groups and for both the employed and the unemployed (Cantor et
al., 2012; Sommers and Kronick, 2012). Other research focused on health care utilization also

46

shows significant increases in health insurance coverage (Sommers et al., 2013; Akosa Antwi et
al., 2013 and 2015; Chua and Sommers, 2014; Barbaresco et al., 2015). This increase in
coverage provides protection to young adults at risk of losing insurance in the absence of the law,
especially for men, unmarried adults, non-students, and those with poor health (Cantor et al.,
2012). Thus the ACA dependent coverage reform has successfully increased health insurance
coverage, which decreases the price of medical care faced by young adults.
B. The Impact of the ACA Dependent Coverage Expansion on Utilization, Health
Outcomes, and Risky Behaviors
One would predict that increases in coverage would lead to overall increases in medical
care access and consumption as a result of reduced medical care prices. Sommers et al. (2013)
show that the ACA dependent coverage expansion reduces delays in getting care and care
foregone due to costs. Akosa Antwi et al. (2015) find that the number of overall (non-birth)
hospital visits as well as inpatient visits associated with a mental health diagnosis increase as a
result of the ACA dependent coverage expansion. 25 The authors do not find evidence of a
noticeable impact on hospital length of stay or number of procedures. Barbaresco et al. (2015)
find an increase in the probability of having a personal doctor as a result of the ACA dependent
coverage expansion. However, they did not find any significant increase with respect to
preventive care utilization. Chua and Sommers (2014) do not find any impact of the ACA
dependent coverage expansion on inpatient or outpatient utilization.
There is less work considering the impact of dependent coverage expansions on health
outcomes or risky behaviors. In terms of health outcomes, Chua and Sommers (2014) show
significant increases in excellent self-reported mental and physical health as a result of the ACA
25

Rather than focusing on gains in coverage, Anderson et al. (2012 and 2014) examine the consequences of young
adults aging off (age 19 and age 23 for students) of their parents insurance plans and find a 61 percent reduction
in inpatient hospital admissions.

47

dependent coverage expansion. Barbaresco et al. (2015) find a statistically significant increase in
self-reported excellent health, but no significant changes in mental health, physical health, or
functional limitations. With respect to risky behaviors, Barbaresco et al. (2015) find mixed
results, with increases in binge drinking, along with decreases in BMI (body mass index).
C. Other Types of Coverage Expansions and Avoidable Hospitalizations
Billings and Teicholz (1990) first developed the concept of using avoidable (or
ambulatory care sensitive (ACS)) hospitalizations as an indirect indicator of problems associated
with primary care quality and access to care. The idea here is that certain hospitalizations could
be avoided if the patient has access to high quality primary care. Thus, this approach allows
researchers to use hospital discharge data, which is readily available, to assess ambulatory care
quality.
Dafny and Gruber (2005) use such an approach to investigate the impact of the Medicaid
expansions of the 1980s and 1990s on low income children made newly eligible for public
coverage using data from the National Hospital Discharge Survey. They find that total
hospitalizations increase significantly as a result of these coverage expansions. A decomposition
of all hospital stays into those that are avoidable versus those that are unavoidable suggests that
the increase for unavoidable hospitalizations is much larger than that for avoidable
hospitalizations. In addition, the increase in avoidable hospitalizations they estimate is not
statistically significant. They take this as evidence that there is an efficiency effect associated
with expanding coverage, but that this efficiency effect is dominated by the access effect.
In the study about the impact of a Medicaid outreach program in California the late 1990s,
Aizer (2007) tests the hypothesis that families responding to outreach efforts will sign their
children up before they get sick, improving their access to outpatient care, and reducing their

48

number of avoidable hospitalizations. Using California Medicaid administrative enrollment and

claims data, she finds that increases in Medicaid take up resulted in lower hospitalization rates
for avoidable conditions, but not others.
Using the HCUP NIS hospital discharge data from 2004 to 2008, KK examine the impact
of the 2006 Massachusetts health insurance reform on avoidable hospitalizations for non-elderly
adults. After controlling for illness severity, the authors estimate a statistically significant,
negative impact of the Massachusetts reform on avoidable hospitalizations. They attributed the
reduction in such hospitalizations to patients with less severe medical problems. Unlike the
previous papers described here, KK employ avoidable hospitalization definitions developed by
the AHRQ specifically for this type of analysis.
Taken as a whole, evidence from the literature suggests that expanding insurance
coverage leads to more primary care utilization. Dafny and Gruber (2005), Aizer (2007), and KK
all hypothesize that this could lead to a reduction in the need for avoidable hospitalizations.
Dafny and Gruber (2005) refer to this as the efficiency effect. On the other hand, an increase in
insurance coverage could lead to more hospitalizations as the price of hospital care falls. Dafny
and Gruber (2005) call this the access effect. Both Aizer (2007) and KK find reductions in
avoidable hospitalizations among the different populations gaining coverage in their studies.
Their findings suggest that the efficiency effect dominates the access effect. Conversely, Dafny
and Gruber (2005) find an increase in avoidable hospitalizations, though not as large as for total
hospitalizations among children gaining Medicaid coverage in the mid-1980s through mid-1990s.
Therefore, it is not obvious which effect would dominate for young adults gaining coverage
through the ACA dependent coverage expansion. The conceptual model described in the next

49

section formalizes this discussion and introduces moral hazard as a third potential channel
through which insurance expansions can impact avoidable hospitalization demand.
III. Conceptual Model
Here I derive a conceptual model of avoidable hospitalizations that will guide my
empirical work. I posit that the probability of an avoidable hospitalization is a function
price of an avoidable hospital stay (
(

) and the consumers health status , so

of the
(

)=

, ). Further, I assume that the price of an avoidable hospital stay is a function of the ACA

mandate or law ( ) and that the consumers health status is a function of their primary care
consumption (

) and their engagement in risky behaviors ( ):

(

Primary care consumption (

)= (

( ), (

, ))

(1)

) is going to depend on the price of primary care (

), which

itself is a function of the ACA ( ), and risky behaviors ( ) are going to depend on the price of
avoidable hospitalizations (

), which itself is also a function of the ACA ( ). Putting this all

together gives me the following equation:

(

)= {

( ),

( ) , (

( )) }

(2)

Since I am interested in the impact of the ACA dependent care coverage expansion on avoidable
hospitalizations, I take the derivative of this function with respect to :
=

+
-

(
-

+
-

(3)

Below each term I include my assumption about its sign based largely on the literature described
in the previous section.
Since the ACA dependent care expansion increased insurance coverage among young
adults, the partial derivatives of the change in health care prices with respect to the law should all
50

be negative, so

and

< 0. The law of demand suggests that as health care prices fall, we

would expect health care consumption to increase, implying that

and

< 0 . These

assumptions imply that the first term on the right hand side of equation (3) is positive and can be
thought of as the access effect. The access effect suggests that as young adults gain insurance
coverage and face lower prices for avoidable hospitalizations, their demand for avoidable
hospitalizations will increase.
The second term on the right hand side formalizes two additional channels through which
a coverage expansion can influence avoidable hospitalization consumption. The first channel is
the efficiency effect. It suggests that as the price of primary care falls, young adults will
consume more primary care (
(

< 0). 26 This in turn is assumed to improve their health

> 0) and reduce their demand for avoidable hospitalizations (

< 0).27 As mentioned in

the previous section, both Aizer (2007) and KK find that the efficiency effect dominates the
access effect since they estimate overall reductions in avoidable hospitalizations among the
populations they study.

26

Several studies have found that insurance expansions increase primary care consumption, including Manning et al.
(1987), Currie and Gruber (1996a), Lichtenberg (2002), Card et al. (2008), and Finkelstein et al. (2012).
27
Does more primary care really improve health? This can be a difficult question to answer with respect to
insurance expansions as such expansions may increase primary care consumption, while at the same time also
potentially increasing risky behavior, and both will affect health outcomes. The literature on the ACA dependent
coverage example discussed in the previous section of this paper suggests mixed findings with respect to health
outcomes. Brook et al. (1983) find that free care improves cholesterol levels, mental and physical health in certain
sub-groups in the RAND health insurance experiment. Several studies suggest that Medicaid expansions reduce
mortality and increase self-assessed overall, mental and physical health, while having no statistically significant
effects on laboratory-measured health outcomes (Currie and Gruber, 1996b; Finkelstein et al., 2012; Sommers et al.,
2012; Baicker et al., 2013). The Medicare program has been estimated to decrease mortality rates for Medicare
inpatients (Card et al., 2009), but no significant impact of Medicare on the mortality rate for the elderly in general
has been found (Finkelstein and McKnight, 2008). Unanimous evidence from the 2006 Massachusetts health
insurance reform shows increases in self-assessed overall, mental and physical health, and decreases in functional
limitations, joint disorders and mortality (Van der Wees et al., 2013; Courtemanche and Zapata, 2014; Sommers et
al., 2014).

51

The other channel captured by the second term on the right hand side of equation (3) can
be thought of as representing ex ante moral hazard (Ehrlick and Becker, 1972) through the term
in two ways. First, a reduction in the price of avoidable hospitalizations could lead to an
increase in risky behaviors such as drinking and smoking. Second, a reduction in the price of
avoidable hospitalizations could lead to a reduction in the demand for health promoting activities,
such as flu vaccinations or smoking cessation program participation. Such behavior suggests
< 0 and I assume that an increase in risky behavior leads to a reduction in health

< 0.

The moral hazard effect would thus be predicted to lead to an increase in avoidable
hospitalizations.
As this channel was not explicitly mentioned in the previous literature on avoidable
hospitalizations, more discussion is warranted. First, is there evidence that reductions in the price
of avoidable hospitalizations lead to increases in risky behaviors and reductions in health
promoting activities? The empirical literature on these topics is mixed. Neither the RAND health
insurance experiment nor the Oregon Medicaid study found a significant impact of insurance
coverage on smoking or body weight (Brook et al., 1983; Finkelstein et al., 2012). While Dave
and Kaestner (2009) find increases in smoking and drinking, and decreasing physical activity,
associated with enrolling in the Medicare program, none of the effects are significant.
Courtemanche and Zapata (2014) find no evidence on smoking or physical activity as a result of
the 2006 Massachusetts health insurance reform, though they do find a significant reduction in
body mass index.28 As mentioned, Barbaresco et al. (2015) find mixed results for risky behaviors
with significant improvement in BMI, but increase in binge drinking.

28

Body mass index is a proxy of poor diets and sedentary lifestyles, and has been broadly used as one of the risky
behaviors in the literature. However, it might not fully satisfy the narrow definition here as it can be affected through

52

As for the impact of risky behaviors on health, Mcginnis and Foege (1993) find in their
influential study that half of the deaths in the United States in 1990 are from external modifiable
risk behaviors. More recent studies by Mokdad et al. (2004 and 2005) also show similar results
for the U.S. in 2000: smoking, diet, physical activity, and drinking are the main risky behaviors
leading to death. Danaei et al. (2009) break dietary behavior into detailed categories and find
high body mass index, physical inactivity, and high blood glucose are the three main risk factors
leading to death, followed by a list of dietary risk factors. In general, the literature supports the
notion that risky behaviors have an impact on health. However, young adults maybe more
immune to the health effects of risky behaviors as compared to the elderly or to children.
Taken together, the evidence on these terms suggests that there may be a moral hazard
effect associated with increased insurance coverage, which would lead to a higher demand for
avoidable hospitalizations. My conceptual model predicts that the efficiency effect would lead to
a reduction in avoidable hospitalizations, while the access effect and the moral hazard effect lead
to increases in avoidable hospitalizations. Thus the overall effect is ambiguous, reinforcing the
need to analyze this issue empirically.
IV. Methodology
I use a difference-in-differences strategy to examine the impact of the ACA dependent
coverage expansion on the prevalence of avoidable hospitalizations among the treatment group
of young adults relative to the control group of slightly older young adults before and after the
mandates implementation in late September of 2010. Because the group targeted by the mandate
is 19-to-25 year olds, most previous studies on the ACA dependent coverage mandate use an age

other channels (such as being suggested or reminded by the doctor during each physician visit) than the pure price
effect of avoidable hospitalizations.

53

range of 19-25 to define their treatment group and typically use older young adults (sometimes
including those as old as 34) as their control group.
The key identifying assumption in any difference-in-differences model is the assumption
that both the treatment and control groups would have experienced the same changes in
outcomes in the absence of the intervention of interest. Slusky (2013) calls into question the
validity of the common trends assumption with respect to labor market outcomes for young
adults in the age rage typically used in the literature. He replicates previous studies with placebo
treatment dates occurring several years prior to the implementation of the mandate and finds
significant effects. This suggests that previous studies may be mistakenly attributing changes
in young adult insurance coverage to the ACA that are actually driven by dynamics in the age
structure of insurance and labor markets. He finds more reliable estimates after reducing the age
bandwidth associated with the treatment group.
Like Barbaresco et al. (2015), I address this concern by defining my treatment group as
young adults aged 23 to 25 and the control group as young adults aged 27 to 29.29 Sluskys
concerns are arguably less important for avoidable hospitalizations than they are for labor market
outcomes, since avoidable hospitalizations are likely less directly impacted by cyclical economic
fluctuations. In addition, narrowing the age bandwidth associated with the treatment and control
groups, as done here and in Barbaresco et al. (2015), should also reduce the impact of any
differential economic shocks. Finally, relative to other studies, I use a longer pre-reform period
(starting from 2002) in my analysis to better test for differences in pre-reform trends between the
treatment and control groups.
Formally, I estimate the following equation:
29

I follow the previous literature and exclude young adults aged 26, as it is difficult to determine whether or not the
mandate is binding for them. It would be a function of their birthdate and the start date of their parents insurance
plan for the year.

54

=
where

is a dummy variable equal to one if hospital discharge is considered an avoidable

in hospital at time . The primary parameter of

hospitalization generated by a patient of age

interest is denoted by
targeted age group.

. It measures the effect of the mandate after implementation on the

is a dummy variable equal to one for any discharge generated by a

patient in the age range of 23-25 (the treatment group).

for any discharge occurring in a time period

is a dummy variable equal to one

that is after the implementation of the ACA

mandate (October 2010 or later). The vector

includes a set of patient demographic

characteristics and a set of risk adjusters to control for patient illness severity. The terms
and

(4)

capture separately age, time, and the hospital fixed effects. Finally

represents the

error term. In my estimation, I use heteroskedasticity-robust standard errors clustered at the

treatment level of the interaction of age-by-time.30 NIS sampling weights, discussed below, are
used in the analysis.
To verify the validity of my findings, I perform several placebo regressions using
treatment dates occurring several years prior to the implementation of the mandate as in Slusky
(2013). I also perform multiple additional robustness checks. The first two checks re-estimate
equation (4) with shorter pre-reform time frames (15 and 23 quarters versus 35) to verify that my
results are not driven by my chosen length of the pre-reform period. The third check excludes the
time period of April 2010 to September 2010 (Q2 2010 Q3 2010), which is the time period
between when the law passed and its effective date, to avoid ambiguity about the treatment status
of hospitalized young adults during this period.

30

The estimated standard errors are similar when they are clustered on age alone. I prefer using an interaction of
age-by-time as it gives more clusters (Angrist and Pischke, Chapter 8).

55

V. Data
The dataset used for this analysis is the Nationwide Inpatient Sample (NIS), which is part
of the Healthcare Cost and Utilization Project (HCUP) administered by the Agency for
Healthcare Research and Quality (AHRQ). Each year of the NIS is a stratified sample of 20
percent of community hospitals in the U.S. and is nationally representative of all community
hospitals.31 If a hospital is sampled in a given year, it provides the universe of its discharges for
that year, regardless of payer. As in KK, I take advantage of the fact that a large fraction of
hospitals are sampled in each year to identify within hospital changes over time.
The NIS is a good data source to examine the impact of health insurance coverage
reforms since it has complete payer information for each discharge. Detailed information on
diagnoses and patients point of admission (directly admitted or transferred from other facilities)
allow me to create indicators for avoidable hospitalizations. One weakness of this data is that it
only consists of hospitalized patients, which may introduce a selection problem with illness
severity into the analysis. I use several patient-level risk adjusters to control for this problem.
The years I use for this analysis range from 2002 to 2011 (the most recent year
available).32 Since the mandate was implemented in late September 2010 and NIS is a quarterly
data, I define the time from the first quarter of 2002 to the third quarter of 2010 as the pre-reform
period, and from the fourth quarter of 2010 to the fourth quarter of 2011 as the post-reform
period. My sample starts with 4,813,849 discharges from the NIS for young adults aged 23-29
over the 2002-2011 period of analysis. After excluding discharges with missing values for key
31

One caveat to note is that not every state participates in this endeavor. By 2011, there are 46 states reporting data
to the HCUP database. Data from Alabama, Delaware, Idaho, and New Hampshire are not available in any year
because they did not provide data to the NIS. Other states report incomplete data. I exclude the states of California,
Maine and Texas from the analysis because detailed age information for patients is not available.
32
The AHRQ redesigned the NIS sampling strategy in 2012. The new NIS is a sample of discharges from all
hospitals participating in HCUP, rather than all discharges from a sample of participating hospitals, as in previous
years. A consistent hospital identifier, allowing researchers to control for hospital fixed effects, will no longer be
available.

56

variables (age, gender, principal diagnosis, quarter, year, and hospital), my sample is reduced to
3,845,814 discharges. A total of 3,363,241 discharges occur in the pre-reform time period and
482,573 occur in the post-reform time period.
The main outcome I consider in this analysis is the classification of a given discharge as
an avoidable hospitalization, which implies that it is a hospitalization for a condition or treatment
which could have been potentially prevented by effective community outpatient / primary care or
other early medical intervention. Thus avoidable hospitalizations serve as a proxy for primary
care quality. Such a hospitalization is also referred to as an ambulatory care sensitive (ACS)
hospital admission.
One issue associated with this literature is that the definition of an avoidable
hospitalization is often ad hoc and can differ from study to study. Given that I am using data
from the AHRQ, I follow KK and use the AHRQ methodology for identifying avoidable
hospitalizations. This methodology identifies twelve separate conditions / treatments considered
to be avoidable for adults, such as an inpatient stay due to dehydration or uncontrolled diabetes.33
The AHRQ provides software that creates flags for each of the twelve conditions / treatments,
which they call Prevention Quality Indicators (PQIs).34
Table 2.1 provides the summary statistics of the pre- and post-reform and corresponding
difference-in-differences calculations for all of the PQI avoidable hospitalization indicators, as
well as AHRQ generated composites for acute and chronic PQIs, and an overall composite. The
definition of the acute composite indicator, PQI 91, is the union of PQI indicators 10, 11, and 12
33

Actually, this methodology identifies fourteen conditions, but I am not considering COPD / asthma admissions
among older adults or low birth weight admissions.
34
The AHRQ software generates these PQIs based on hospital discharge data by using complex algorithms.
Essentially, the indicators first look for specific principal diagnoses, then exclude certain discharges based on their
secondary and tertiary diagnoses. Transfers from other facilities are excluded to avoid double-counting. A diagnosis
of pregnancy, if necessary, is also excluded in certain PQIs. For more information on the AHRQ PQI methodology,
see: http://www.qualityindicators.ahrq.gov/Modules/PQI_TechSpec.aspx

57

(dehydration, bacterial pneumonia, and urinary tract infections). Similarly, the definition of the
chronic composite indicator, PQI 92, is the union of PQI indicators 1, 3, 7, 8, 13, 14, 15, and 16
(short-term and long-term diabetes complications, hypertension, congestive heart failure, angina,
uncontrolled diabetes, adult asthma, and lower-extremity amputation). Thus it includes all PQIs
except the previously defined acute indicators and PQI 2 (perforated appendix), because it has a
different denominator. Finally, the overall PQI indicator (PQI 90) is defined as the union of all of
the individual indicators except PQI 2.
The first row of table 2.1 suggests that in the pre-reform time period, the probability of a
discharge among a young adult in the treated group being avoidable is 3.48 percent while the
probability of a discharge among a young adult in the control group being avoidable is 3.55
percent. There is also a slightly lower probability of a discharge being chronic avoidable for
young adults in the treated group than in the control group (1.81 vs 1.94 percent) in the prereform time period. For the acute PQI composite, the probability of a discharge being acute
avoidable among the treated group is 1.67 percent, while it is 1.61 percent in the control group.
For the twelve individual PQI indicators, most discharges have a slightly higher probability of
being avoidable in the control group in the pre-reform time period, except short-term diabetes
(PQI 1) and urinary tract infections (PQI 12).
The simple difference-in-differences calculations presented in the last column of table 2.1
compare the changes of the mean probability for the treatment relative to control group in the
pre- and post-reform periods, showing statistically significant increases in the overall PQI
composite, the chronic PQI composite, as well as the PQIs for short-term diabetes complications
(PQI 1), congestive heart failure (PQI 8), dehydration (PQI 10), angina without a procedure (PQI
13), and uncontrolled diabetes (PQI 14). The calculations also show statistically significant

58

decreases in PQI 7 and 16. This is suggestive evidence that the ACA dependent coverage
expansion may have led to an increase in avoidable hospitalizations.
Figure 1 shows trends in the probability that a given discharge is an overall, acute or
chronic avoidable hospitalization separately for the treatment and control groups. The figures
show similar trends for both groups before mandate, indicating that time-variant changes in
observables and unobservables may not differ substantially between the two groups. This
provides further support for implementing a difference-in-differences analysis.
In order to isolate the impact of the ACA dependent coverage mandate, I include in my
regression analysis a set of demographic control variables. These are dummy variables for each
year of age, gender, race/ethnicity, and patients zip code in income quartile. In addition, I
include the quarterly state unemployment rate, from the Bureau of Labor Statistics, to control for
state level economic conditions. Following KK, I also utilize a set of risk adjusters to control for
patient disease severity. These risk adjusters include the number of diagnoses on the discharge
record, AHRQ comorbidity dummies for different diseases, All-Patient Refined Diagnosis
Related Groups (APR-DRGs) classification, the APR-DRG severity of illness score, and the
APR-DRG risk of mortality score.35 All the risk adjusters are designed to measure some level of
illness severity and are included in my discharge level regression.
Table 2.2 shows the pre-reform means and standard deviations for the demographic
controls for the young adult discharges in the sample. Within both the treatment and the control
group, the discharges are evenly distributed across the age categories. A larger share of
discharges is generated by females (81.3 percent) than males, with similar percentages in both

35

The number of diagnoses is calculated by counting the number of diagnoses on each discharge record. The AHRQ
comorbidity dummies provide 29 categories of disease comorbidity (i.e. for congestive heart failure: 1 represents
comorbidity and 0 shows comorbidity is not present). The APR-DRG related measures, developed by 3M, are used
to classify patients according to their degree of potential mortality and illness severity.

59

the treatment group and the control group. As for race and ethnicity, discharges from whites
make up a slightly lower (40.9 percent vs. 43.5 percent) share for treatment group as compared
to the control group. For the patients zip code income quartile, discharges associated with the
age group 23-25 have a higher share (33.5 percent) in the two lowest quartiles, as compared to
30.6 percent among discharges from the age group 27-29.
VI. Results
A. Average Effects of the ACA Dependent Coverage Expansion on the Probability of
Avoidable Hospitalizations
Table 2.3 provides the results of difference-in-differences estimation of the baseline
model representing equation (4) (left panel) and a similar model including patient risk adjusters
(right panel). The baseline model suggests a 0.12 percentage point increase in the probability of a
discharge being avoidable, which represents a 3.4 percent increase compared to the baseline rate
of avoidable hospitalizations.36 The composite indicator for chronic avoidable hospitalizations
also shows a significant increase of 0.09 percentage points, which represents a 5 percent increase.
The coefficient on the composite acute indicator, although not significant, is also positive.
Additionally, table 2.3 lists results for each individual PQI indicator. Among the twelve
individual indicators in the baseline model, five of them suggest statistically significant increases
in a discharge being associated with that particular avoidable admission (short-term diabetes
complications, congestive heart failure, dehydration, angina, and uncontrolled diabetes); two
exhibit statistically significant reductions (hypertension and lower-extremity amputation) and the
remaining five have no statistical significance.

36

Compared to the pre-reform treatment mean of 3.48 percent, the increase of 0.12 percentage point represents an
increase of 3.4 percent.

60

To control for potential changes in the patient population in the post-reform time period, I
estimate the same model with risk adjusters, where I use severity of disease to control for
observable changes in the health status of the patient pool. These results are presented in the
right panel of table 2.3. The estimates are similar to those generated by the baseline model, with
slightly higher effects associated with overall (4.9 percent increase vs. 3.4 percent) and chronic
avoidable hospitalizations (7.7 percent increase vs. 5 percent). This suggests that the illness
severity of the inpatient population for young adults did not change much after the ACA mandate,
which may due to the fact that young adults are relatively healthy in general.
B. Placebo Tests
In order to test the validity of the difference-in-differences results presented in the
previous sub-section, I estimate a series of four placebo tests that use artificial effective dates
within the pre-reform period as in Slusky (2013). Following previous studies (Antwi Akosa et al.,
2014 and 2015; Barbaresco et al., 2015) which use a five-year period for their primary analyses,
I use five-year windows pre-reform for my placebo tests spanning 2005-2009, 2004-2008, 20032007, and 2002-2006.37 In my baseline model, there are five quarters in the post-reform time
period, so I also use five quarters as the length of my artificial post-reform time period in each
placebo test (e.g. the fourth quarter of 2008 is the start of the artificial post-reform time period
for the 2005-2009 placebo test). I estimate a specification similar to my baseline model for all of
the PQIs in each of the four placebo tests.
Table 2.4 reports the estimates from these tests. Fifteen PQI regressions in each of the
four sets of placebo tests generate a total of 56 regressions. Theoretically, a small number of
significant results are expected due to the large number of regressions. Around one estimate is
37

In unreported placebo tests (available upon request), I estimate another five placebo tests with varying time
windows of 2002-2009 (8 years), 2002-2008 (7 years), 2002-2007 (6 years), 2002-2006 (5 years), and 2002-2005 (4
years). The results are similar in terms of the number of significant estimates.

61

expected to be significant at the 1 percent level, three at the 5 percent level, and six at the 10
percent level by chance. The number of significant results reported in table 2.4 is 0 at the 1
percent level, four (6.7 percent) at the 5 percent level, and eight (13.3 percent) at the 10 percent
level. Note that one particular PQI, PQI 13 (angina), accounted for three of the eight significant
results. Dropping PQI 13 from the definitions of the overall PQI avoidable hospital indicator and
the chronic composite indicator does not lead to major changes in my primary results. Overall,
these placebo tests suggest that my primary difference-in-differences approach is sound and there
does not appear to be any sustained differential pre-reform trends between the treatment and
control groups. Moreover, these placebo test results also suggest that the standard errors, which
are clustered at the age-by-time level, are not meaningfully understated.
C. Robustness Checks
Here I describe the results of multiple robustness checks that are presented in table 2.5.
For ease of comparison, the first column of table 2.5 re-states my baseline results. Columns two
and three restrict the period of analysis to 2007-2011 and 2005-2011 respectively. In each case
the estimated impact of the ACA dependent coverage expansion on the likelihood that a young
adult discharge is avoidable is very similar in terms of magnitude and statistical significance.
The coefficient estimate in the baseline model suggests a 0.12 percentage point increase, while
the coefficient estimate in the 2007-2011 (2005-2011) model suggests a 0.11 (0.12) percentage
point increase. The results are similar for both the chronic and acute composite PQI indicators.
This suggests my baseline results are not being driven by the length of the pre-reform period.
The next robustness check, presented in column four, drops the time period between the
passage of the ACA and its dependent coverage expansion implementation date, which I define

62

as the second and third quarters of 2010. As above, making this change does not impact the
coefficient estimates in a major way.
D. Heterogeneity Tests
Having verified the validity of my empirical model and estimated the average effects of
the ACA dependent coverage expansion, I next present the results of models that allow for
heterogeneous effects for different sub-groups in my sample. There may be differences by
gender or race in response to gaining insurance coverage. In addition, differences in
socioeconomic status may also lead to different responses. Tables 6 and 7 present the results
from heterogeneity regressions based on gender, race and patients zip code income quartile.
The first two columns of table 2.6 illustrate differences by gender. These results suggest
that the statistically significant increases in the probability of an overall avoidable hospitalization
(PQI 90) or a chronic avoidable hospitalization (PQI 92) in my baseline model are being driven
by young females, rather than young males. Young men do statistically significantly reduce their
probability of a hospitalization for hypertension (PQI 7) and extremity amputation (PQI 16) after
gaining coverage, but increase their probability of a hospitalization for heart failure (PQI 8).
The next three columns of table 2.6 present differences by race. Black, Hispanic, Asian,
Native American and other races compose 28 percent of the sample and are grouped together as
non-white. The remaining sample is classified as either unknown race (30 percent) or white (42
percent). These results suggest that the statistically significant increases in the probability of an
overall avoidable hospitalization (PQI 90) or a chronic avoidable hospitalization (PQI 92) in my
baseline model are being driven by whites, rather than non-whites or those with unknown race.
Although I find no statistically significant impact on acute avoidable hospitalizations (PQI 91) in

63

my baseline model, table 2.6 suggests that the ACA dependent coverage expansion lead to an
increase in the probability of having such a hospitalization for non-whites.
Table 2.7 presents heterogeneity model results based on patients zip code income
quartile.
These results suggest that the statistically significant increase in the probability of an overall
avoidable hospitalization (PQI 90) in the baseline model are being driven by patients coming
from zip codes with income that fall in the second or third income quartile of the distribution.
The increase in the probability of a chronic avoidable hospitalization (PQI 92) in the baseline
model is being driven by patients coming from zip codes with income that fall in the third
income quartile of the distribution. Taken together, this heterogeneity analysis suggests that there
are important differences by gender, race, and income in response to gaining insurance through
the ACA dependent coverage mandate.
VII. Discussion
The overall increase in the probability of avoidable hospitalizations suggested by my
empirical analysis implies that the access effect and the moral hazard effect dominate the
efficiency effect for young adults gaining coverage through the ACA dependent coverage
expansion. This is broadly consistent with the finding in Antwi Akosa et al. (2015) that ACA
dependent coverage expansion increases non-birth hospital admissions and admissions
associated with a mental health diagnosis. I find some evidence of an efficiency effect for young
adult avoidable hospitalizations as there are two individual indicators (hypertension and
extremity amputation) that show reductions in probability after the ACA mandate. This echoes
the results found in Dafny and Gruber (2005) for children gaining Medicaid coverage. Among
those children there was some evidence of an efficiency effect, but this was dominated by the

64

access effect. On the other hand, Aizer (2007) finds that when eligible, but not enrolled children
formally enroll in Medicaid coverage in California they experience a reduction in avoidable
hospitalizations. This would suggest the efficiency effect dominates.38
Relative to my results, studies from the Massachusetts health insurance expansion tell a
different story for older (non-elderly) adults. KK find that the Massachusetts reform leads to a
reduction in the probability of avoidable hospitalizations, which they implicitly attribute to the
efficiency effect dominating the access effect. This suggests that the older adults targeted by the
reform responded by increasing their primary care consumption, thus reducing their rate of
avoidable hospital stays. The difference in findings for young adults from ACA expansion and
older (non-elderly) adults from Massachusetts reform may due to several potential reasons:

Information or experience: Gaining health insurance coverage may lead to reductions

in avoidable hospitalizations (i.e. the efficiency effect dominates), but that requires the
newly insured to seek out and receive appropriate primary care. Young adults gaining
coverage through the ACA dependent coverage expansion may not have enough
experience with the health care system to successfully find such primary care services.
Older adults are more likely to have this needed experience.

Risk attitudes: Additionally, these older adults may be more risk averse than young
adults, as they may realize that their overall health is no longer as good as when they

38

Why do the results from Dafny and Gruber (2005) and Aizer (2007) regarding avoidable hospitalizations for
children newly enrolled in Medicaid vary? One possible explanation is that Dafny and Gruber (2005) focus on
children made newly eligible for Medicaid, while Aizer (2007) focuses on already eligible children who are now
formally taking up Medicaid coverage. Presumably, children made newly eligible for Medicaid did not have a
previous source of coverage for hospital or primary care. On the other hand, families of children who are eligible,
but not formally enrolled in Medicaid may understand that hospital care would still be covered by Medicaid, as the
hospital likely has experience assisting such families in the Medicaid enrollment process. This is less likely to be
true with respect to primary care. Therefore, one could consider eligible, but not formally enrolled children as
having conditional hospital coverage but not conditional primary care coverage. Thus the children analyzed in
Aizer (2007) experienced a greater increase in access to primary care as compared to hospital care. This increase in
primary care access could explain why avoidable hospitalizations for this particular group of children fall.

65

were younger. Older adults may also need to protect themselves more diligently so that
certain infectious disease (such as the flu) will not affect their family members. Therefore,
even though the price of hospital care decreases due to expansions in insurance coverage,
non-elderly adults do not want to face the risk of being hospitalized and so make sure
they consume the necessary primary care.

Income constraints and Moral Hazard: For financial reasons, young adults may be
more likely than the older adults to forgo insurance coverage and instead focus on lower
cost interventions such as flu vaccines and over-the-counter medications. However,
receiving insurance coverage alleviates the financing constraint, and as a result, young
adults may engage in more risky behavior or invest less in their health, such as increasing
binge drinking (Barbaresco et al., 2015). In other words, the ex ante moral hazard effect
of obtaining coverage may be stronger for young adults than other adults.
On the other hand, dependent health insurance coverage may also increase young adults

disposable income, as some of them may no longer have to pay their own insurance premium.
They may use this extra income to consume goods with adverse health consequences, such as
cigarettes and alcohol. Barbaresco et al. (2015) show an increase in risky drinking; increases in
drinking may lead to heart disease and diabetes in the long-run.
This discussion illustrates the benefits of using a conceptual model to think about how the
impact of gaining coverage might differ for individuals of different ages. While my results might
seem at first glance to contradict the results from Massachusetts, there are several plausible
reasons why we might expect young adults to respond differently to a gain in insurance coverage
than older adults.

66

VIII. Conclusion
A typical hospitalization may be characterized as an unavoidable because there is nothing
that could have been done medically to avoid the stay, such as suffering a major injury in a car
accident. In this paper I investigate whether or not there were changes in the probability of
having an avoidable hospitalization one that could have been prevented by the receipt of timely
and appropriate primary medical care among young adults gaining health insurance coverage
through ACA dependent coverage expansion which was implemented in September 2010.
Though several previous studies have examined the impact of coverage expansions on hospital
utilization, there are many reasons why we might expect young adults to potentially respond
differently than older adults or children. To answer this question I use HCUP NIS hospital
discharge data and AHRQ avoidable hospitalization definitions to estimate a difference-indifferences model with a narrow age bandwidth of age 23-25 as the treatment group and age 2729 as the control group. The results shown in the baseline model for the entire sample indicate
increases in the probability of having any avoidable hospitalization as well as the chronic
composite, but no clear effects on the acute composite index.
Specifically, the ACA dependent coverage mandate leads to an increases in the
probability of PQI 1 (short-term diabetes), PQI 8 (congestive heart failure), acute PQI 10
(dehydration), PQI 13 (angina), and PQI 14 (uncontrolled diabetes). At the same time, I estimate
decreases in the probability PQI 7 (hypertension) and PQI 16 (lower-extremity amputation).
Controlling for patient illness severity does not lead to major changes in these results. I then
utilize several placebo regressions with pre-reform periods to validate the model with a narrow
age range treatment group. Next I implement four robustness checks to confirm the effects
shown in the baseline model are not driven by my choice of the length of the pre-reform period

67

in my analysis. Finally, I estimate the model on sub-samples of different gender, race, and zip
code income quartiles. There are important differences by gender, race, and income in response
to the ACA dependent coverage mandate.

68

Table 2.1 Means and Standard Deviations for Outcome Variables

Quality Indicators
Overall Prevention Quality Indicators
PQI 90 Overall Composite

Pre-reform Period
Treatment
Control
(Ages 23-25)
(Ages 27-29)

Post-reform Period
Treatment
Control
(Ages 23-25)
(Ages 27-29)

Difference-inDifference

0.0348 (0.1832)

0.0355 (0.1851)

0.0381 (0.1916)

0.0368 (0.1883)

0.0021 (0.0012)*

0.0167 (0.1281)

0.0161 (0.1260)

0.0165 (0.1274)

0.0154 (0.1230)

0.0006 (0.0006)

0.0181 (0.1332)
0.0194 (0.1379)
PQI 92 Chronic Composite
Individual Component Measures of Prevention Quality Indicators
0.0079 (0.0885)
0.0067 (0.0814)
PQI 01 Diabetes short-term comp.

0.0216 (0.1455)

0.0214 (0.1448)

0.0015 (0.0008)*

0.0108 (0.1033)

0.0085 (0.0917)

0.0011 (0.0005)**

PQI 91 Acute Composite

PQI 02 Perforated appendix

0.1717 (0.3771)

0.1748 (0.3798)

0.1763 (0.3812)

0.1853 (0.3887)

-0.0059 (0.0127)

PQI 03 Diabetes long-term comp.

0.0020 (0.0446)

0.0030 (0.0550)

0.0027 (0.0523)

0.0039 (0.0622)

-0.0001 (0.0002)

PQI 07 Hypertension

0.0005 (0.0222)

0.0009 (0.0305)

0.0005 (0.0234)

0.0012 (0.0348)

-0.0002 (0.0001)**

PQI 08 Congestive heart failure

0.0009 (0.0303)

0.0015 (0.0390)

0.0009 (0.0307)

0.0013 (0.0356)

0.0003 (0.0001)**

PQI 10 Dehydration

0.0041 (0.0642)

0.0043 (0.0655)

0.0034 (0.0585)

0.0032 (0.0561)

0.0005 (0.0002)**

PQI 11 Bacterial pneumonia

0.0053 (0.0728)

0.0058 (0.0757)

0.0059 (0.0752)

0.0060 (0.0772)

0.0001 (0.0005)

PQI 12 Urinary tract infection

0.0072 (0.0847)

0.0061 (0.0776)

0.0074 (0.0857)

0.0062 (0.0786)

0.00001 (0.0003)

PQI 13 Angina without procedure

0.0001 (0.0096)

0.0002 (0.0134)

0.0001 (0.0081)

0.0001 (0.0101)

0.0001 (0.00003)*

PQI 14 Uncontrolled diabetes

0.0008 (0.0277)

0.0009 (0.0299)

0.0008 (0.0280)

0.0007 (0.0272)

0.0002 (0.0001)**

PQI 15 Asthma in younger adults

0.0059 (0.0767)

0.0062 (0.0782)

0.0058 (0.0756)

0.0057 (0.0753)

0.0003 (0.0005)

PQI 16 Lower-extremity amputation 0.00003 (0.0053)

0.0001 (0.0082)

0.00002 (0.0042)

0.0001 (0.0113)

-0.0001 (0.00003)**

1,743,153

225,861

256,712

--

Sample Size

1,620,088

Notes: Means are reported, with standard deviations in parentheses. Standard errors, heteroskedasticity-robust and clustered at the age-by-time level, are in
parentheses for difference-in-differences calculations. NIS sampling weights are used. *** indicates the difference-in-differences is significant at the 1% level;
** 5%; * 10%.

69

Table 2.2 Pre-reform Means and Standard Deviations for Control Variables
Control Variables

Total
(Ages 23-29)

Age dummies (age=23 is omitted)

Age=24
0.161 (0.367)
Age=25
0.164 (0.371)
Age=27
0.173 (0.378)
Age=28
0.173 (0.378)
Age=29
0.173 (0.378)
Female
0.813 (0.390)
Race/ethnicity dummies (non-Hispanic white is omitted)
Black
0.130 (0.336)
Hispanic
0.089 (0.285)
Asian
0.018 (0.131)
Native American
0.006 (0.079)
Other than black, Hispanic, Asian, Native, or white
0.036 (0.185)
Unknown Race
0.299 (0.458)
Patients Zip Code in Income Quartile dummies (First (Lowest) is omitted)
Second Income Quartile
0.154 (0.361)
Third Income Quartile
0.133 (0.340)
Fourth Income Quartile
0.099 (0.299)
Unknown Income
0.448 (0.497)
State Unemployment Rate
6.059 (2.056)
Notes: Means are reported, with standard deviations in parentheses. NIS sampling weights are used.

70

Treatment
(Ages 23-25)

Control
(Ages 27-29)

0.334 (0.472)
0.341 (0.474)
---0.813 (0.390)

--0.333 (0.471)
0.333 (0.471)
0.334 (0.471)
0.812 (0.391)

0.139 (0.346)
0.094 (0.291)
0.015 (0.121)
0.007 (0.081)
0.035 (0.184)
0.301 (0.459)

0.121 (0.327)
0.085 (0.279)
0.020 (0.140)
0.006 (0.077)
0.036 (0.186)
0.297 (0.457)

0.158 (0.365)
0.127 (0.333)
0.084 (0.277)
0.454 (0.498)
6.041 (2.048)

0.150 (0.357)
0.139 (0.346)
0.113 (0.317)
0.442 (0.497)
6.077 (2.064)

Table 2.3 Difference-in-Differences Estimates of Effects of ACA Dependent Coverage Mandate on Quality Indicators
Quality Indicators

Baseline Model

with Risk Adjusters

Pre-reform mean (treated group)

0.0012 (0.0005)**

0.0017 (0.0005)***

0.0348

0.0004 (0.0004)

0.0003 (0.0004)

0.0167

0.0009 (0.0004)**

0.0014 (0.0004)***

0.0181

Overall Prevention Quality Indicators

PQI 90 Overall Composite
PQI 91 Acute Composite
PQI 92 Chronic Composite

Individual Component Measures of Prevention Quality Indicators

PQI 01 Diabetes short-term

0.0009 (0.0003)***

0.0010 (0.0003)***

0.0079

PQI 02 Perforated appendix

-0.0021 (0.0116)

0.0003 (0.0088)

0.1717

PQI 03 Diabetes long-term

-0.0003 (0.0002)

-0.00003 (0.0002)

0.0020

PQI 07 Hypertension

-0.0003 (0.0001)***

-0.0003 (0.0001)***

0.0005

PQI 08 Heart failure

0.0002 (0.0001)**

0.0003 (0.0001)***

0.0009

PQI 10 Dehydration

0.0004 (0.0002)**

0.0005 (0.0002)***

0.0041

PQI 11 Bacterial pneumonia

0.00002 (0.0002)

-0.0001 (0.0002)

0.0053

PQI 12 Urinary tract infection

-0.00004 (0.0002)

-0.0001 (0.0002)

0.0072

PQI 13 Angina

0.00005 (0.00002)*

0.0001 (0.00003)*

0.0001

PQI 14 Uncontrolled diabetes

0.0002 (0.0001)**

0.0002 (0.0001)***

0.0008

0.0001 (0.0002)

0.0003 (0.0002)

0.0059

-0.0001 (0.00003)***

-0.0001 (0.0002)***

0.00003

PQI 15 Asthma (younger)

PQI 16 Extremity amputation
Sample Size
a

3,845,814

3,812,595

1,620,088

Notes: For PQI 2, the sample size is 48,748. For PQI 2, the sample size is 48,275. *** indicates the difference-in-difference is significant at the 1% level; **
5%; * 10%. Standard errors, heteroskedasticity-robust and clustered at the age-by-time level, are in parentheses. All regressions include the controls plus age,
hospital and time fixed effects. NIS sampling weights are used.

71

Table 2.4 Placebo Regressions

2005-2009

2004-2008

2003-2007

2002-2006

Treatment 2007 Q4

Treatment 2006 Q4

Treatment 2005 Q4

Treatment 2004 Q4

PQI 90

0.0007 (0.0006)

0.0004 (0.0004)

0.0004 (0.0006)

0.0003 (0.0006)

PQI 91

0.0002 (0.0004)

0.0001 (0.0004)

-0.0003 (0.0004)

-0.0003 (0.0004)

PQI 92

0.0006 (0.0005)

0.0003 (0.0004)

0.0007 (0.0004)*

0.0005 (0.0004)

PQI 01

0.0002 (0.0003)

-0.0001 (0.0003)

-0.00001 (0.0003)

-0.0001 (0.0002)

PQI 02

0.0128 (0.0120)

-0.0272 (0.0120)**

0.0007 (0.0113)

0.0054 (0.0118)

PQI 03

0.0002 (0.0002)

-0.0003 (0.0002)**

0.0002 (0.0002)

0.0001 (0.0001)

PQI 07

-0.0001 (0.0001)

-0.00002 (0.0001)

0.0001 (0.0001)

0.00003 (0.0001)

PQI 08

0.0001 (0.0001)

0.0002 (0.0001)

0.0001 (0.0001)

0.0002 (0.0001)

PQI 10

0.0001 (0.0002)

-0.0002 (0.0002)

-0.0002 (0.0002)

0.00004 (0.0002)

PQI 11

0.00002 (0.0003)

0.0003 (0.0002)

0.0001 (0.0002)

-0.0001 (0.0002)

PQI 12

0.00001 (0.0003)

0.0001 (0.0002)

-0.0003 (0.0003)

-0.0002 (0.0003)

PQI 13

0.0001 (0.00002)**

0.0001 (0.00003)*

-0.00001 (0.00003)

0.0001 (0.00004)*

PQI 14

-0.0001 (0.0001)

-0.0001 (0.0001)

0.0002 (0.0001)*

-0.000003 (0.0001)

PQI 15

0.0002 (0.0003)

0.0006 (0.0003)**

0.0002 (0.0002)

0.0003 (0.0003)

PQI 16

0.00000 (0.00002)

0.00003 (0.00002)

0.00001 (0.00002)

-0.00001 (0.00002)

Quality Indicators

Notes: *** indicates significant at the 1% level; ** 5%; * 10%. Standard errors, heteroskedasticity-robust and clustered at the age-by-time level, are in
parentheses. All regressions include the controls plus age, hospital and time fixed effects. NIS sampling weights are used.

72

Table 2.5 Robustness Checks

Drop periods

Quality Indicators

Baseline Model

2007-2011

2005-2011

PQI 90

0.0012 (0.0005)**

0.0011 (0.0006)**

0.0012 (0.0005)**

0.0012 (0.0005)**

PQI 91

0.0004 (0.0004)

0.0004 (0.0004)

0.0005 (0.0004)

0.0004 (0.0004)

PQI 92

0.0009 (0.0004)**

0.0007 (0.0004)*

0.0008 (0.0004)*

0.0008 (0.0004)**

PQI 01

0.0009 (0.0003)***

0.0009 (0.0003)***

0.0009 (0.0003)***

0.0009 (0.0003)***

PQI 02

-0.0021 (0.0116)

-0.0021 (0.0121)

-0.0027 (0.0118)

-0.0044 (0.0118)

PQI 03

-0.0003 (0.0002)

-0.0003 (0.0002)

-0.0003 (0.0002)*

-0.0003 (0.0002)

PQI 07

-0.0003 (0.0001)***

-0.0003 (0.0001)***

-0.0003 (0.0001)***

-0.0003 (0.0001)***

PQI 08

0.0002 (0.0001)**

0.0002 (0.0001)

0.0002 (0.0001)*

0.0002 (0.0001)**

PQI 10

0.0004 (0.0002)**

0.0005 (0.0002)***

0.0005 (0.0002)***

0.0004 (0.0002)**

PQI 11

0.00002 (0.0002)

-0.0001 (0.0002)

-0.0002 (0.0002)

0.00004 (0.0002)

PQI 12

-0.00004 (0.0002)

-0.00002 (0.0003)

-0.00000 (0.0002)

-0.00001 (0.0002)

PQI 13

0.00005 (0.00002)*

0.00001 (0.00003)

0.00002 (0.00003)

0.00005 (0.00002)*

PQI 14

0.0002 (0.0001)**

0.0002 (0.0001)**

0.0002 (0.0001)**

0.0002 (0.0001)**

PQI 15

0.0001 (0.0002)

0.0001 (0.0003)

0.0002 (0.0003)

0.0001 (0.0002)

PQI 16

-0.0001 (0.00003)***

-0.0001 (0.00003)***

-0.0001 (0.00003)***

-0.0001 (0.00003)***

Sample Size

3,845,814

1,975,809

2,749,374

3,645,578

2010 Q2 - 2010 Q3

Notes: *** indicates significant at the 1% level; ** 5%; * 10%. Standard errors, heteroskedasticity-robust and clustered at the age-by-time level, are in
parentheses. All regressions include the controls plus age, hospital and time fixed effects. NIS sampling weights are used.

73

Table 2.6 Heterogeneity by Gender and Race

Quality
Indicators

Female

Male

White

Non-white

Unknown
Race

PQI 90

0.0018 (0.0005)***

0.0017 (0.0015)

0.0020 (0.0007)***

0.0015 (0.0010)

0.0008 (0.0015)

PQI 91

0.0004 (0.0004)

0.0001 (0.0007)

-0.0003 (0.0005)

0.0011 (0.0005)**

-0.0001 (0.0011)

PQI 92

0.0014 (0.0004)***

0.0017 (0.0014)

0.0023 (0.0005)***

0.0004 (0.0008)

0.0009 (0.0009)

PQI 01

0.0011 (0.0003)***

0.0007 (0.0011)

0.0013 (0.0004)***

0.0007 (0.0005)

0.00001 (0.0008)

PQI 02

0.0071 (0.0121)

-0.0040 (0.0103)

0.0017 (0.0114)

0.0081 (0.0128)

-0.0128 (0.0245)

PQI 03

0.00001 (0.0002)

-0.0001 (0.0006)

0.0002 (0.0002)

-0.0006 (0.0004)*

0.0007 (0.0003)**

PQI 07

-0.0002 (0.0001)**

-0.0005 (0.0002)**

-0.0001 (0.0001)

-0.0004 (0.0002)**

-0.0003 (0.0002)

PQI 08

0.0001 (0.0001)

0.0013 (0.0004)***

0.0001 (0.0001)

0.0008 (0.0002)***

0.0001 (0.0002)

PQI 10

0.0006 (0.0001)***

-0.0005 (0.0005)

0.00002 (0.0002)

0.0007 (0.0003)***

0.0011 (0.0004)**

PQI 11

-0.0002 (0.0002)

0.0006 (0.0005)

-0.00001 (0.0002)

0.0003 (0.0003)

-0.0005 (0.0005)

PQI 12

0.00002 (0.0003)

-0.0001 (0.0004)

-0.0003 (0.0003)

0.0004 (0.0004)

-0.0007 (0.0006)

PQI 13

0.00004 (0.0000)**

0.0001 (0.0001)

0.00003 (0.0000)

0.0001 (0.0001)

0.0001 (0.0000)***

PQI 14

0.0002 (0.0001)***

0.0002 (0.0003)

0.0001 (0.0001)

0.0003 (0.0002)*

0.0003 (0.0001)**

PQI 15

0.0003 (0.0002)

0.0001 (0.0006)

0.0007 (0.0003)**

-0.0004 (0.0005)

0.0001 (0.0004)

PQI 16

-0.00003 (0.0000)

-0.0002 (0.0001)**

-0.00004 (0.0000)*

-0.0001 (0.0000)*

-0.0001 (0.0001)*

Sample Size

3,096,019

716,576

1,646,963

1,100,320

1,065,312

Notes: *** indicates significant at the 1% level; ** 5%; * 10%. Standard errors, heteroskedasticity-robust and clustered at the age-by-time level, are in
parentheses. All regressions include the controls plus age, hospital and time fixed effects. NIS sampling weights are used.

74

Table 2.7 Heterogeneity by Patients Zip Code Income Quartile

Quality
Indicators

First (Lowest)
Income Quartile

Second
Income Quartile

Third
Income Quartile

Fourth
Income Quartile

Unknown
Income Quartile

PQI 90

0.0001 (0.0010)

0.0022 (0.0011)**

0.0035 (0.0011)***

0.0009 (0.0015)

0.0008 (0.0034)

PQI 91

0.00003 (0.0008)

0.0014 (0.0008)*

0.0008 (0.0007)

-0.0002 (0.0009)

-0.0015 (0.0018)

PQI 92

0.0001 (0.0009)

0.0008 (0.0008)

0.0028 (0.0009)***

0.0011 (0.0009)

0.0023 (0.0026)

PQI 01

0.0012 (0.0006)*

-0.0006 (0.0006)

0.0013 (0.0007)*

0.0017 (0.0006)***

0.0022 (0.0017)

PQI 02

-0.0169 (0.0207)

0.0097 (0.0163)

0.0415 (0.0162)**

-0.0364 (0.0220)

0.0484 (0.0769)

PQI 03

-0.0008 (0.0004)**

-0.0002 (0.0004)

0.0011 (0.0003)***

-0.0003 (0.0004)

0.0003 (0.0008)

PQI 07

-0.0004 (0.0002)**

-0.0002 (0.0002)

-0.0003 (0.0001)**

0.0001 (0.0001)

-0.0005 (0.0004)

PQI 08

0.0008 (0.0002)***

0.0004 (0.0002)**

-0.0002 (0.0002)

-0.0002 (0.0002)

-0.0011 (0.0009)

PQI 10

0.0005 (0.0003)

0.0006 (0.0004)*

0.0005 (0.0003)

0.0001 (0.0004)

0.0009 (0.0009)

PQI 11

-0.0005 (0.0004)

0.0006 (0.0003)*

-0.00002 (0.0004)

0.0002 (0.0005)

-0.0013 (0.0014)

PQI 12

0.00003 (0.0005)

0.0002 (0.0005)

0.0003 (0.0004)

-0.0005 (0.0007)

-0.0011 (0.0016)

PQI 13

0.0001 (0.0001)

-0.0001 (0.0001)

0.00003 (0.00003)

-0.00004 (0.0001)

0.0003 (0.0001)**

PQI 14

-0.0001 (0.0002)

0.0004 (0.0002)**

0.0004 (0.0001)***

-0.0001 (0.0001)

0.0006 (0.0008)

PQI 15

-0.0007 (0.0005)

0.0011 (0.0005)**

0.0004 (0.0004)

-0.0001 (0.0005)

0.0004 (0.0015)

PQI 16

-0.0001 (0.0001)**

0.00000 (0.0000)

-0.00004 (0.0000)

-0.0001 (0.0000)

0.00001 (0.0000)

Sample Size

698,365

639,567

565,782

409,215

1,499,666

Notes: *** indicates significant at the 1% level; ** 5%; * 10%. Standard errors, heteroskedasticity-robust and clustered at the age-by-time level, are in
parentheses. All regressions include the controls plus age, hospital and time fixed effects. NIS sampling weights are used.

75

Figure 2.1 Trends in Prevention Quality Indicators by Age Group

(a) Overall Prevention Quality Indicator

Age23~25

3
Q

20 Q 3
03
Q
1
20 Q 3
04
Q
1
20 Q 3
05
Q
1
20 Q 3
06
Q
1
20 Q 3
07
Q
1
20 Q 3
08
Q
1
20 Q 3
09
Q
1
20 Q 3
10
Q
1
20 Q 3
11
Q
1

20

02
Q

.0 3

.0 3 5

.0 4

.0 4 5

Overall Prevention Quality Indicator (PQI90)

Age27~29

(b) Acute Prevention Quality Indicator

Age23~25

3
Q

20 Q 3
03
Q
1
20 Q3
04
Q
1
20 Q3
05
Q
1
20 Q 3
06
Q
1
20 Q3
07
Q
1
20 Q 3
08
Q
1
20 Q 3
09
Q
1
20 Q 3
10
Q
1
20 Q 3
11
Q
1

20

02

.0 1 4

.0 1 6

.0 1 8

.0 2

Acute Prevention Quality Indicator (PQI91)

Age27~29

(c) Chronic Prevention Quality Indicator

Age23~25

3
Q

20 Q 3
03
Q
1
20 Q3
04
Q
1
20 Q3
05
Q
1
20 Q 3
06
Q
1
20 Q3
07
Q
1
20 Q 3
08
Q
1
20 Q 3
09
Q
1
20 Q 3
10
Q
1
20 Q 3
11
Q
1

20

02

.0 1 5

.0 2

.0 2 5

Chronic Prevention Quality Indicator (PQI92)

Age27~29

76

CHAPTER III
Health Insurance and Traffic Fatalities for Young Adults
I. Introduction
On September 23rd, 2010, the dependent coverage expansion of the Patient Protection and
Affordable Care Act (ACA) was implemented to increase health insurance coverage for young
adults aged 19 to 25. Prior to this expansion, young adults in this age range would commonly
age off of their parents insurance plans and often become uninsured due to a lack of other
sources of insurance.39
Several studies examine the impact of ACA dependent coverage expansion on health
insurance coverage; these studies employ different datasets and consistently find a statistically
significant increase in coverage for young adults (Cantor et al., 2012; Sommers and Kronick,
2012; Sommers et al., 2013; Akosa Antwi et al., 2013 and 2015; Chua and Sommers, 2014;
Barbaresco et al., 2015). Less explored in the literature however is moral hazard, a potential
unintended consequence associated with such an expansion in coverage. The purpose of
insurance is to protect the insured from financial problems due to large losses (such as disease,
accidents, loss of valuables etc.). The theory of moral hazard predicts that when the potential
costs from these losses are borne, in whole or in part, by others, the insured have a tendency to
take more risks. Increases in risk-taking behavior associated with a health insurance expansion
could happen in two ways: one is through reductions in the consumption of preventive care that
an individual might otherwise consume in order to avoid costly hospitalizations; the other is
through increases in risky behaviors, such as smoking, binge drinking, or over eating. Excessive
drinking may also be associated with drunk driving, which can lead to fatal traffic accidents.

39

For more on this policy see: http://www.hhs.gov/healthcare/rights/youngadults/index.html

77

Concerns about drunk driving leading to fatal traffic accidents are particularly important
for young adults. According to Insurance Institute for Highway Safety (IIHS), although young
drivers are less likely to drink and drive than adults, their crash and fatality risks are higher if
they do, due to their relative inexperience associated with both drinking and driving.40 All fifty
states and the District of Columbia have imposed Graduated Driver Licensing (GDL) programs
in an attempt to reduce teen drivers driving risk through enhancing driving restrictions. Of
course, reckless driving behavior, especially drinking and driving, is not easily regulated.41
In the earlier study of the impact of the ACA young adult health insurance coverage
expansion on health outcomes (Barbaresco, Courtemanche and Qi, 2015), we find some potential
evidence of moral hazard among newly insured young adults, that an increase in health insurance
coverage leads to an increase in risky drinking behavior. As described the above, another
potential channel for moral hazard associated with new insurance coverage would be through
increases in reckless, including alcohol-impaired driving. This, in turn, could lead to more traffic
accidents and more traffic fatalities.
What do we know about the rate of fatal traffic accidents for young adults during our
timeframe of interest? According to the National Vital Statistics Reports, young adults in the age
group of 15 to 24 have the highest number of deaths (20 percent of all ages) caused by motor
vehicle accidents. 42 Also, motor vehicle crashes are the leading cause of death (15.9 percent of
all causes) for young adults in the same age group (Centers for Disease Control and Prevention,
2011). Among all the fatal accidents, alcohol related crashes make up 36 percent for all ages
(National Highway Traffic Safety Administration, 2011).

40

For more information, see: http://www.iihs.org/iihs/topics/t/teenagers/topicoverview.

For more information about statewide Graduated driver licensing program, see:
http://www.iihs.org/iihs/topics/laws/graduatedlicenseintro?topicName=teenagers.
42
The report is at: http://www.cdc.gov/nchs/data/nvsr/nvsr61/nvsr61_04.pdf, Table7.
41

78

In this paper, I estimate the causal relationship between health insurance coverage and
traffic fatalities among young adults; that is, whether or not the increase in health insurance
coverage for young adults through the ACA dependent coverage expansion leads to more overall
traffic fatalities and more alcohol-related traffic fatalities. My primary results suggest that for
young adults (aged 20) who just gained health insurance from the ACA dependent coverage
expansion increased their risky driving behavior, leading to more traffic accidents and even more
traffic fatalities. The magnitude associated with the increase in traffic accidents is smaller when I
restrict attention to alcohol-related accidents. The rest of this chapter is organized as follows:
Section II presents a literature review of traffic fatalities; section III describes the data used in
this study; section IV illustrates the methodology employed in my empirical work; section V
presents my primary results, and section VI concludes the chapter.
II. Literature Review
There is a broad literature investigating the impact of specific types of insurance on
related types of injuries. Cohen and Dehejia (2004) find that an increase in the share of drivers
with auto insurance increases traffic fatalities and Bolduc et al. (2002) find that increase of
generosity in workers compensation insurance increase injuries related to work. To the best of
my knowledge, there is no previous research directly investigating the impact of health insurance
coverage on traffic fatalities. As will be discussed in more detail below, there are, however,
established strands of the literature exploring the impact of alcohol consumption, prices changes,
and Body Mass Index (BMI) on fatal crashes. The ACA dependent coverage expansion has been
shows to influence each of these factors, so it could in turn have an impact of fatal crashes.
One strand of the literature debates on the impact of beer taxes and minimum drinking
age laws (MDAL) on traffic fatalities. Chaloupka et al. (1993) show that increased beer taxes has
79

the largest impact on reducing youth fatality (an 11.5 percent reduction), followed by mandatory
administrative license laws (a 9 percent reduction). The authors also indicate that MDAL reduces
total fatalities by about 5 to 6 percent. Saffer and Grossman (1987) conclude that the elasticity of
the motor vehicle fatality rate to the real beer tax is about 20 for young adults. This suggests that
an increase in beer taxes could reduce the youth death rate. Similarly, Ruhm (1996) finds that,
compared to relatively small impacts from other regulations (i.e. MDAL), increases in beer taxes
cause larger reductions in youth fatalities. However, Dee (1999) suggests that beer taxes have a
relatively small and statistically insignificant impact on teen drinking after controlling for crossstate heterogeneity; while the implementation of MDAL actually leads to reductions in heavy
teen drinking by 8 percent and reductions in traffic fatalities by at least 9 percent. This debate
suggests that in my analysis of the impact of expansions in health insurance coverage I should
control for the money cost of alcohol, as well as any non-pecuniary costs faced by young adults
when attempting to acquire alcohol. In addition, I should control for potential cross-state
heterogeneity, as state laws prior to the ACA may weaken the impact of the federal policy
changes.
Another strand of the literature looks at how income changes impact traffic fatalities
among young adults. Adams et al. (2012) show that an increase of 10 percent in the minimum
wage has a positive correlation of 5 to 10 percent with alcohol-related accidents for teen drivers.
Grabowski and Morrisey (2004) find that a 10-cent decrease in gasoline prices leads to an
increase in motor vehicle fatalities over a 2-year period. They also find that the effect is larger
for higher-risk young adult drivers. By looking at the changes in state gasoline taxes, Grabowski
and Morrisey (2006) suggest that plausibly exogenous increases in state gasoline taxes are
related to fewer traffic fatalities. This strand of literature implies that income effects, such as an
80

increase in income when a young adult substitutes insurance coverage they pay for with costless
coverage through their parents plan, are related to fatal traffic crashes.
A third strand of the literature considers the relationship between obesity and driving
behavior. Anderson et al. (2012) show that commercial motor vehicle operators with higher BMI
were more likely to be in a subsequent accident. Simmons and Zlatoper (2010) find that during
2005, accident fatalities per mile traveled was positively associated with a states obesity
prevalence. Dunn and Tefft (2013) investigate the relationship between BMI and traffic fatalities
among young adults and find that obesity tends to make the body less inebriated and helps
decrease traffic fatalities related to alcohol consumption. As in the earlier study of the impact of
the ACA dependent coverage expansion on health outcomes (Barbaresco, Courtemanche, and Qi,
2015), we find that ACA dependent coverage expansion helped improve BMI among young
adults, which may reduce traffic fatalities.
A final strand of the literature I consider focuses on the impact of statewide Graduated
Driver Lisensing (GDL) programs on young adults traffic fatalities. Dee et al. (2005) and
Morrisey et al. (2006) find GDL regulations reduce traffic fatalities among 15-17 year olds.
Morrisey and Grabowsky (2010) find that good GDL programs reduce overall traffic fatalities,
as well as driver fatalities for the policys targeted age groups (teenagers from 15-17 and 18-20
age groups). Therefore, controlling for the type of state GDL program is also important when
evaluating other policies related to traffic fatalities of young adults.
There are also many studies that consider show a mix of factors mentioned above and
tend to give us an ambiguous prediction for the impact of each factor. Gallet (2007) finds that
higher income leads to greater alcohol consumption but lowers the risk of being overweight.
Courtemanche (2010) finds a negative relationship between gasoline prices and BMI as higher
81

gasoline prices encourage more walking and less dinning out. As the mechanism of maintaining
sobriety by heavier weight may differ by personal physical body function, it may not be very
efficient to be considered as a factor that can be easily influenced by policy.
III. Data
A. Fatality Analysis Reporting System
Data on fatal vehicle crashes are obtained from the Fatality Analysis Reporting System
(FARS) of the National Highway Traffic Safety Administration (NHTSA). The FARS is a
census of all motor vehicle trafc accidents that result in a fatality for either occupants or nonmotorists. It includes detailed information on the characteristics of the vehicles, drivers,
occupants and non-occupants involved in the crash. Because I am interested in the relationship
between insurance coverage of young adults and traffic fatalities, I restrict attention to accidents
resulting in a fatality caused by young drivers. As discussed in the following sub-sections,
policies such as Minimum Drinking Age Laws and Zero Tolerance Laws associated with
drinking, as well as Graduated Driver Licensing programs targeted at young adults below age 21,
tried to reduce traffic fatalities caused by younger drivers. Thus, in this study, I mainly focus on
young drivers aged 20 and below as the primary treatment group. Following Morissey and
Grabowski (2010), by state, year, and age group of interest, I count the number of traffic
accidents resulting in a fatality as well as the total number of fatalities associated with these
accidents.43 These counts serve as the primary dependent variables in my analysis, which is
focused on 2008-2013 in order to have an equal number of years before and after the ACA
expansion of dependent coverage.

43

Following previous literature, I exclude Alaska, Hawaii and the District of Columbia due to the fewer
observations.

82

The results of blood alcohol concentration (BAC) tests for those involved in traffic
accidents are sometimes not reported. The FARS attempts to impute these missing values and
adopted a new method for doing so in 2001.44 Because I use imputed values of BAC test results
in my analysis, I do not include any data prior to 2001. This insures the imputation method for
the BAC test results is constant throughout every year in my sample. Since young adults may be
more easily involved in an alcohol-related crash, any accident record with a BAC greater than 0
will be counted as alcohol-related accident.45 This is consistent with the enforcement of Zero
Tolerance Laws for young drivers below age 21.
B. Graduated Driver Licensing Program
Graduated driver licensing (GDL) laws have been established to help reduce traffic
accidents among teen drivers. GDL regulation varies across states and across three distinct
licensing stages. These three licensing stages are: learner stage (minimum entry age, mandatory
permit holding period, and minimum amount of supervised driving), intermediate stage with
unsupervised driving (unsupervised nighttime driving prohibition and restriction on passengers),
and unrestricted stage (restrictions lifted age).46
By using a standardized classification system created by the Insurance Institute for
Highway Safety (IIHS), one can characterize GDL laws into four groups (Good, Fair, Marginal,
and Poor) to evaluate the restrictiveness of the laws. Those criteria include the required length of
holding a learners permit (usually six months), restrictions on unsupervised nighttime driving
(usually 10pm to 5am), the number of teen passengers (usually no more than one) in the car, and

44

For more discussion about new imputation methodology, see: http://www-nrd.nhtsa.dot.gov/Pubs/809-450.pdf.

Zero Tolerance Laws in different states have different criteria regarding acceptable BAC levels, ranging from 0.00
to 0.02 percent. For more information, see: http://dui.findlaw.com/dui-laws-resources/underage-dui-zero-tolerancelaws.html.
46
For detailed regulations by states, see:
http://www.iihs.org/iihs/topics/laws/graduatedlicenseintro?topicName=teenagers.
45

83

the minimum age (usually age 17) until the restrictions are lifted. The most restrictive states in
terms of teen driving are placed in the good category, while state with the least restrictions are
placed in the poor category. Previous studies (Morrisey et al., 2006; Morrisey and Grabowski,
2010) find that states with GDL programs in the good and fair categories have fewer young
adult motor vehicle fatalities. In this study, I classify state GDL programs by year according to
the IIHS criteria and use this to control a states young adult driving environment.
Following the literature, other control variables I employ include beer tax rates, gasoline
prices, state unemployment rates, and state total population. The beer tax data were obtained
from the Tax Foundation.47 This source provides tax rates (dollars per gallon) for beer, as well as
other alcohol, for each state in each year. Annual average regular grade gasoline wholesale/resale
prices (dollars per gallon) by refiners were obtained from the U.S. Energy Information
Administration.48 Both the tax and price data were adjusted for inflation based on the 2013
annual CPI. Annual average state unemployment rates were obtained from the Bureau of Labor
Statistics (BLS). Total population by state, year and age were constructed using data from
Current Population Survey (CPS).49
IV. Methodology
I use a difference-in-differences approach to identify the impact of the ACA dependent
coverage expansion on traffic accidents and fatalities. Equation (1) below described the
47

For further information, see: http://taxfoundation.org/article/state-sales-gasoline-cigarette-and-alcohol-tax-rates.

I use annual average wholesale/resale gasoline prices since annual average retail gasoline prices were not
available for years after 2010. Resale prices are usually slightly lower than retail prices as the intermediary
businesses earn the price differences, and the price differences should stay stable over years. In the sample,
wholesale/resale prices and retail prices for gasoline before 2011 have the same time trends. This suggests I can use
wholesale/resale gasoline prices as a proxy for retail gasoline prices.
49
Specific age information was not available from the Census Bureau. To validate the effectiveness of using the
CPS, I constructed the total number of observation with weights from CPS in the corresponding age group from the
ACS (American Community Survey). The CPS with weights has around 0.5% fewer total individuals in each age
group as compared to the ACS. Since total population at each age serves as a state-year-age group level control,
consistently lower number of population make it plausible to use the CPS to construct total population in each
estimated cell.
48

84

empirical model I estimate with the data described in the previous section:
=

where

+
+

+
+

+
+

is accident or fatality counts in age group , state , and year ;

mandates on the treatment over control group;

+
,

(1)
captures the effect of

is a dummy variable that indicates the

gender of the driver, the pedestrian, or the bicyclist that caused the fatal accident;
represents the gasoline price for each state in each year;
each state in each year;

represents categories of restrictiveness of the Graduated Driver

Licensing program in force in the state in the relevant year;

unemployment rate for each state in each year;
each year.

and

represents the beer tax for

is the annual average

is the annual total population in each state in

control for age, year and state fixed effects separately for the

drivers and the accidents they caused, and

is the error term. Standard errors are clustered at

the treatment level of the interaction of age-by-time.50

As mentioned above, for the main analysis, I use data from 2008 to 2013 to allow for
three years of pre-reform data (2008-2010) and three years of post-reform data (2011-2013).51 In
order to cleanly estimate the causal impact of the ACA dependent coverage expansion, the
treatment group I analyze consists of young drivers aged 20, and the control group is the young
drivers aged 18. Age 19 was excluded from the analysis as it is hard to tell whether they have
been dropped from their parents plan due to their birth date and the renewal dates of their parents
health insurance plan. Teenagers aged 15 to 17 were excluded for two reasons. First, they should

50

If clustered by age alone, there are only two clusters in the regression. I prefer using an interaction of age-by-time
as it gives more clusters (Angrist and Pischke, Chapter 8).
51
I treat 2010 as a pre-treatment year since insurance companies may not expand the dependent coverage until
parents for whom the reform is binding renew their health insurance coverage, normally the beginning of the next
year.

85

not be impacted by the ACA dependent coverage expansion, since they were likely already
eligible for insurance through their parents employer or for public coverage through the CHIP
program.52 Second, most states set the minimum driving age to be no earlier than age 16, and
some states dont lift supervised driving restrictions until age 17 or 18.53 Since we cannot tell
from the FARS which drivers are college students for whom the ACA dependent coverage
mandate is not binding, I assume all drivers are not students. This should result in lower bound
estimates of the impact of the reform.
The fundamental identifying assumption associated with difference-in-differences
analysis is the parallel trends assumption for the treatment and control groups in the pretreatment period. Figure 1 shows the trends for two outcomes traffic accident counts and
traffic fatality counts - over the entire period of study for both sets of treatment and control
groups. Figure 2 is the corresponding figures for alcohol-related traffic accident counts and
traffic fatality counts. Trends for the pre-treatment periods for each set of treatment and control
groups are parallel to each other, which validates the use of the difference-in-differences
approach in this study.
Table 3.1 shows means and standard errors for all traffic accidents and fatalities (panel I),
as well as alcohol-related traffic accidents and fatalities (panel II) for both the treatment and
control group. Panel I shows that both accident counts and fatality counts for treatment group
(age 20) have similar means as in the control group (age 18) in the pre-treatment period. After
the ACA dependent coverage mandate, means in treatment group are much higher than the
52

See Marton (2007) and Marton and Talbert (2010) for more on the CHIP program.
Eight states (Arkansas, Florida, Georgia, Missouri, New Jersey, Texas, Virginia, and Washington) plus
Washington, D.C. set their full privilege minimum driving age at 18. Seven states (Connecticut, Illinois, Maryland,
Massachusetts and Nevada; New York and Pennsylvania) have specific requirements regarding their full privilege
minimum driving age, such as night driving restrictions or restrictions related to driver education completion, but
also set the full privilege minimum driving age at 18. Thus age 17 will not be included in the control group of the
main analysis, but will be included in the control groups of my robustness checks.
53

86

means in control group. This suggests significant net increases in accidents and fatalities
associated with the ACA dependent coverage mandate according to the simple difference-indifferences calculations presented in the last column. Panel II shows similar patterns when we
restrict attention to alcohol-related traffic accidents and fatalities only, though with smaller
magnitudes. The results in Table 3.1 reflect the trends in Figures 1 and 2. The simple differencein-differences calculation shows a simple, unadjusted pre and post comparison. My regression
models will provide more precise estimates by controlling for other confounding factors that
might influence traffic accidents and fatalities.
V. Results
Table 3.2 shows regression results for both the full sample of traffic accidents and the
alcohol-related-only sub-sample. Results from panel I and II are for full sample, and panel III
and panel IV are for alcohol-related-only sub-samples. The first column of each table is the
baseline regression with years 2008-2013 for age group 20 vs age 18. Results from column 1 of
panel I show that after the implementation of the ACA dependent coverage expansion, young
adults aged 20 experience a 4.4 percentage point (17.0 percent) increase in traffic accidents and a
5.6 percentage point (19.4 percent) increase in traffic fatalities.
The next three columns are robustness checks. To show that the estimates are not affected
by the chosen length of pre-treatment periods, column 2 and 3 of panel I are estimated with a
longer pre-treatment period, one is from 2005, and the other is from 2001, the first available year
in the sample. Column 4 excludes 2010, as the ACA dependent coverage mandate was
implemented in late 2010 and some insurance companies may have enrolled dependents in the
last three quarters of the year in compliance with the reform. In addition, 2007 was included to
maintain the same number of year pre and post reform. Panel II is another set of robustness
87

checks for the full sample analysis. Column 1 restates the baseline estimation from column 1 of
panel I. Columns 2 to 4 are estimations with broader age groups. When these robustness checks
are applied to the full sample of traffic accidents and fatalities, the results remain stable.
Panel III and IV present similar baseline models and robustness checks for the alcoholrelated traffic accident sub-sample. Here I would expect a smaller impact of the ACA dependent
coverage expansion among those aged 20 since the legal drinking age in the United States is 21.
As expected, the coefficient estimates in panels III and IV are smaller in magnitude than the
estimates given in panels I and II. Results from column 1 of panel III show that young adults
aged 20 have a 1.3 percentage point (11.9 percent) increase in traffic accidents and a 1.6
percentage point (12.6 percent) increase in traffic fatalities. The corresponding figures for
broader age groups are given in Figures 3.3-3.8.
VI. Conclusion
Young adults aged 20 who are newly insured by the ACA expansion of dependent
coverage may be more likely to engage in risky behavior, such as reckless driving, and even
drinking and driving, than those aged 18, who were already covered by other types of health
insurance. This could be due to a reduction in their health insurance spending increasing their
disposable income and allowing them to buy more alcohol and / or to drive more miles than
before. Gaining dependent insurance coverage through a parent may also induce these young
adults to drop out of college and/or be more willing to accept a part-time job that does not offer
health insurance. This could potentially increase the amount of driving they do by providing
them with more leisure time and thus increase the potential for traffic accidents.
One caveat of using FARS for younger adults who may not be affected by the ACA
dependent coverage expansion due to their student status is that FARS does not have education
88

information for drivers. Thus, the estimates provided in this study only show a lower bound of
the impact of ACA dependent coverage mandate as I am assuming all the younger drivers in my
sample are not college students.
This study focuses on the younger adults who just obtained health insurance coverage
from ACA dependent coverage expansion and found an increase in traffic accidents and fatalities
for them. Older young adults who finished their education (aged 23-25) may behave differently
from those younger peers when gaining health insurance from their parents. Future study will
examine the impact of ACA dependent coverage expansion on older young adults.

89

Table 3.1 Unadjusted Difference-in-Differences Estimates of the Impact of the ACA

Dependent Coverage Expansion on Traffic Accidents and Fatalities
Panel I: Treatment (age 20) VS Control (age 18) All Traffic Accidents
Outcome
Pre-treatment Periods (08-10)
Post-treat periods (11-13)
Variables
Treat T1
Control C1
Treat T1
Control C1
(age 20)
(age 18)
(age 20)
(age 18)
Accident Count
25.71
25.35
23.38
18.98
(1.62)
(1.45)
(0.17)
(0.42)
Fatality Count
28.83
29.38
26.08
21.42
(1.79)
(1.78)
(0.17)
(0.44)

DifferenceinDifferences
4.04*
(2.22)
5.20*
(2.57)

Panel II: Treatment (age 20) VS Control (age 18) Alcohol-Related Accidents Only
Outcome
Pre-treatment Periods (08-10)
Post-treat periods (11-13)
DifferenceinVariables
Treat T1
Control C1
Treat T1
Control C1
Differences
(age 20)
(age 18)
(age 20)
(age 18)
Accident Count
10.97
8.89
10.44
7.17
1.19
(0.72)
(alcohol-related)
(0.57)
(0.34)
(0.23)
(0.16)
Fatality Count
12.38
10.24
11.69
8.08
1.47*
(0.71)
(alcohol-related)
(0.53)
(0.34)
(0.27)
(0.17)
Notes: Means are reported. Standard errors, heteroskedasticity-robust and clustered by age-by-time, are in
parentheses. *** indicates the difference-in-differences is significant at the 1% level; ** 5% level; * 10% level.

90

Table 3.2 Multivariate Difference-in-Differences Estimates of the Impact of the ACA

Dependent Coverage Expansion on Traffic Accidents and Fatalities
Panel I: All Traffic Accidents - Treatment (age 20) VS Control (age 18)
Outcome
2008-2010 VS
2005-2010 VS
2001-2010 VS
Variables
2011-2013
2011-2013
2011-2013
Accident
4.368***
4.611***
5.425***
Count
(0.278)
(0.281)
(0.389)
Fatality
5.586***
5.606***
6.219***
Count
(0.276)
(0.375)
(0.457)
N
562
848
1,231

2007-2009 VS
2011-2013
4.381***
(0.276)
5.518***
(0.278)
561

Panel II: All Traffic Accidents - Pre (years 2008-2010) VS Post (years 2011-2013)
Outcome
Age 20 VS
Age 20 VS
Age 20-21 VS
Age 20-22 VS
Variables
Age 18
Age 17-18
Age 17-18
Age 16-18
Accident
4.368***
3.282***
3.466***
3.223***
Count
(0.278)
(0.552)
(0.534)
(0.605)
Fatality
5.586***
3.986***
4.332***
3.968***
Count
(0.276)
(0.685)
(0.695)
(0.729)
N
562
843
1,124
1,676
Panel III: Alcohol-Related Accidents Only - Treatment (age 20) VS Control (age 18)
Outcome
2008-2010 VS
2005-2010 VS
2001-2010 VS
2007-2009 VS
Variables
2011-2013
2011-2013
2011-2013
2011-2013
Accident
1.302***
1.128***
1.459***
0.783*
Count
(0.340)
(0.356)
(0.398)
(0.371)
Fatality
1.560***
1.277***
1.640***
0.951*
Count
(0.361)
(0.394)
(0.475)
(0.458)
N
536
816
1,191
536
Panel IV: Alcohol-Related Accidents Only - Pre (years 2008-2010) VS Post (years 2011-2013)
Outcome
Age 20 VS
Age 20 VS
Age 20-21 VS
Age 20-22 VS
Variables
Age 18
Age 17-18
Age 17-18
Age 16-18
Accident
1.302***
0.981**
0.873**
0.748**
Count
(0.340)
(0.391)
(0.352)
(0.342)
Fatality
1.560***
1.051**
1.041**
0.941**
Count
(0.361)
(0.426)
(0.431)
(0.419)
N
536
777
1,054
1,540
Notes: *** indicates significant at the 1% level; ** 5% level; * 10% level. Standard errors, heteroskedasticity-robust
and clustered by age-by-time, are in parentheses. All regressions include the controls plus age, state, and time fixed
effects.

91

Figure 3.1 Traffic Accident / Fatality Counts for Young Adults Aged 20 VS 18
Traffic Fatality

20

20

25

30

30

40

35

40

50

Traffic Accident

2001

2003

2005

2007

2009

Age20

2011

2013

2001

2003

Age18

2005

2007

2009

Age20

fit line

2011

2013

Age18

fit line

Figure 3.2 Alcohol-Related Traffic Accident / Fatality Counts for Young Adults Aged 20
VS 18
Alcohol-Related Traffic Fatality

10

10

15

15

20

20

Alcoho-Related Traffic Accident

2001

2003

2005

2007

Age20

2009

2011

2013

2001

2003

Age18

2005

2007

Age20

fit line

fit line

92

2009

2011
Age18

2013

Figure 3.3 Traffic Accident / Fatality Counts for Young Adults Aged 20 VS 17-18

10

15

20

20

25

30

30

40

35

50

Traffic Fatality

40

Traffic Accident

2001

2003

2005

2007

2009

Age20

2011

2013

2001

2003

Age17~18

2005

2007

2009

Age20

fit line

2011

2013

Age17~18

fit line

Figure 3.4 Alcohol-Related Traffic Accident / Fatality Counts for Young Adults Aged 20
VS 17-18
Alcohol-Related Traffic Fatality

10

10

15

15

20

20

Alcohol-Related Traffic Accident

2001

2003

2005
Age20

2007

2009

2011

2013

2001

2003

Age17~18

2005
Age20

fit line

fit line

93

2007

2009

2011

2013

Age17~18

Figure 3.5 Traffic Accident / Fatality Counts for Young Adults Aged 20-21 VS 17-18

10

15

20

20

25

30

30

40

35

50

Traffic Fatality

40

Traffic Accident

2001

2003

2005

2007

2009

Age20-21

2011

2013

2001

2003

Age17~18

2005

2007

2009

Age20-21

fit line

2011

2013

Age17~18

fit line

Figure 3.6 Alcohol-Related Traffic Accident / Fatality Counts for Young Adults Aged 2021 VS 17-18

25

Alcohol-Related Traffic Fatality

10

10

15

15

20

20

Alcohol-Related Traffic Accident

2001

2003

2005
Age20-21

2007

2009

2011

2013

2001

2003

Age17~18

2005
Age20-21

fit line

fit line

94

2007

2009

2011

2013

Age17~18

Figure 3.7 Traffic Accident / Fatality Counts for Young Adults Aged 20-22 VS 16-18

10

10

20

20

30

30

40

50

Traffic Fatality

40

Traffic Accident

2001

2003

2005

2007

2009

Age20-22

2011

2013

2001

2003

Age16~18

2005

2007

2009

Age20-22

fit line

2011

2013

Age16~18

fit line

Figure 3.8 Alcohol-Related Traffic Accident / Fatality Counts for Young Adults Aged 2022 VS 16-18

25

Alcohol-Related Traffic Fatality

10

10

15

15

20

20

Alcohol-Related Traffic Accident

2001

2003

2005
Age20-22

2007

2009

2011

2013

2001

2003

Age16~18

2005
Age20-22

fit line

fit line

95

2007

2009

2011

2013

Age16~18

REFERENCES
Acemoglu, D., Finkelstein, A., and Notowidigdo, M. J. (2013). Income and Health Spending:
Evidence from Oil Price Shocks. Review of Economics and Statistics, 95, 1079-1095.
Adams, S., Blackburn, M. L., & Cotti, C. D. (2012). Minimum Wages and Alcohol-related
Traffic Fatalities among Teens. Review of Economics and Statistics, 94(3), 828-840.
Aizer, A. (2007). Public Health Insurance, Program Take-up, and Child Health. Review of
Economics and Statistics, 89(3), 400-415.
Akosa Antwi, Y., Moriya, A. S., & Simon, K. (2013). Effects of Federal Policy to Insure Young
Adults: Evidence from the 2010 Affordable Care Act's Dependent-Coverage Mandate.
American Economic Journal: Economic Policy, 5(4), 1-28.
Akosa Antwi, Y., Moriya, A. S., & Simon, K. (2015). Access to Health Insurance and the Use of
Inpatient Medical Care: Evidence from the Affordable Care Act Young Adult Mandate.
Journal of Health Economics, 39, 171-187.
Anderson, M., Dobkin, C., and Gross, T. (2012). The Effect of Health Insurance Coverage on
the Use of Medical Services. American Economic Journal: Economic Policy 4, 1-27.
Anderson, M., Dobkin, C., and Gross, T. (2014). The Effect of Health Insurance on Emergency
Department Visits: Evidence from an Age-Based Eligibility Threshold. Review of
Economics and Statistics 96, 189-195.
Anderson, J. E., Govada, M., Steffen, T., Thorne, C., Varvarigou, V., Kales, S., & Burks, S.
(2012). Health Behavior and Accident Risk: Obesity Is Associated with the Future Risk
of Heavy Truck Crashes among Newly Recruited Commercial Drivers. IZA Discussion
Paper #6408.
Angrist, J. D. and Pischke, J. S. (2008). Mostly Harmless Econometrics: An Empiricists
Companion. Princeton, NJ: Princeton University Press.
Apouey, B. and Clark, A. E. (2014). Winning Big but Feeling No Better? The Effect of Lottery
Prizes on Physical and Mental Health. Health Economics, forthcoming.
Baicker, K., Taubman S. L., Allen, H. L., Bernstein, M., Gruber, J. H., Newhouse, J. P.,
Schneider, E. C., Wright, B. J., Zaslavsky, A. M., and Finkelstein, A. N. (2013). The
Oregon Experiment Effects of Medicaid on Clinical Outcomes. New England Journal
of Medicine, 368, 1713-1722.
Barbaresco, S., Courtemanche, C. J., & Qi, Y. (2015). Impacts of the Affordable Care Act
Dependent Coverage Provision on Health-Related Outcomes of Young Adults. Journal of
Health Economics, 40, 54-68.
96

Billings, J., & Teicholz, N. (1990). Uninsured Patients in District of Columbia Hospitals. Health
Affairs, 9(4), 158-165.
Blum, A. B., Kleinman, L. C., Starfield, B., & Ross, J. S. (2012). Impact of State Laws that
Extend Eligibility for Parents' Health Insurance Coverage to Young Adults. Pediatrics,
129(3), 426-432.
Bolduc, D., Fortin, B., Labrecque, F., & Lanoie, P. (2002). Workers' Compensation, Moral
Hazard and the Composition of Workplace Injuries. Journal of Human Resources, 37(3),
623-652.
Brook, R. H., Ware, J. E., Rogers, W. H., Keeler, E. B., Davies, A., Donald, C., Goldberg, G. A.,
Lohr, K., Masthay, P. C., and Newhouse, J. P. (1983). Does Free Care Improve Adults
Health? Results From a Randomized Controlled Trial. New England Journal of Medicine,
309, 1426-1434.
Callahan, T. S., & Cooper, W. O. (2005). Uninsurance and Health Care Access among Young
Adults in the United States. Pediatrics, 116(1), 88-95.
Cameron, A. C., Gelback, J.B., and Miller, D.L. (2008). Bootstrap-Based Improvements for
Inference with Clustered Errors. Review of Economics and Statistics, 90, 414-427.
Cantor, J. C., Monheit, A. C., Belloff, D., DeLia, D., & Koller, M. (2010). Dependent Coverage
Expansions: Estimating the Impact of Current State Policies. Issue Brief.
Cantor, J. C., Monheit, A. C., DeLia, D., and Lloyd, K. (2012). Early Impact of the Affordable
Care Act on Health Insurance Coverage of Young Adults. Health Services Research,
47(5), 1773-1790.
Card, D., Dobkin, C., and Maestas, N. (2008). The Impact of Nearly Universal Insurance
Coverage on Health Care Utilization: Evidence from Medicare. American Economic
Review, 98, 2242-2258.
Card, D., Dobkin, C., and Maestas, N. (2009). Does Medicare Save Lives? Quarterly Journal of
Economics, 124, 597-636.
Cawley, J. (2004). The Impact of Obesity on Wages. Journal of Human Resources, 39, 451474.
Cawley, J., & Ruhm, C. (2011). The Economics of Risky Health Behaviors. In M. V. Pauly, T. G.
McGuire & P. P. Barros (Eds.), Handbook of Health Economics (Vol. 2, pp. 95-199):
Elsevier.
Cawley, J., Moran, J., and Simon, K. (2010). The Impact of Income on the Weight of Elderly
Americans. Health Economics, 19, 979-993.
97

Centers for Disease Control and Prevention (2009) QuickStats: Percentage of Young Adults
Aged 18-29 Years with Selected Chronic Conditions, by Sex, National Health Interview
Survey, 58(25), 699.
Centers for Disease Control and Prevention (2012). Binge Drinking is Bigger Problem than
Previously Thought. Available
http://www.cdc.gov/media/releases/2012/p0110_binge_drinking.html, accessed 12/13/14.
Centers for Disease Control and Prevention (2014). Past Weekly Surveillance Reports.
Available http://www.cdc.gov/flu/weekly/pastreports.htm, accessed 12/12/14.
Chaloupka, F. J., Saffer, H., & Grossman, M. (1993). Alcohol Control Policies and Motorvehicle Fatalities. Journal of Legal Studies, 22, 161-186.
Chua, K.P. and Sommers, B.D. (2014). Changes in Health and Medical Spending among Young
Adults under Health Reform. Journal of the American Medical Association, 311, 24372439.
Cohen, A., & Dehejia, R. (2004). The Effect of Automobile Insurance and Accident Liability
Laws on Traffic Fatalites. Journal of Law and Economics, 47(2), 357-393.
Collins, S. R., Robertson, R., Garber, T., & Doty, M. M. (2012). Young, Uninsured, and in Debt:
Why Young Adults Lack Health Insurance and How the Affordable Care Act Is Helping.
Commonwealth Fund, Issue Brief(14), 1-24.
Courtemanche, C. J. (2010). A Silver Lining? Reconnection between Gasoline Prices and
Obesity. Economic Inquiry, 49, 935-957.
Courtemanche, C. and Zapata, D. (2014). Does Universal Coverage Improve Health? The
Massachusetts Experience. Journal of Policy Analysis and Management, 33, 36-69.
Courtemanche, C., Heutel, G., and McAlvanah, P. (forthcoming). Impatience, Incentives, and
Obesity. Economic Journal.
Courtemanche, C., Pinkston, J., and Stewart, J. (2014). Adjusting Body Mass for Measurement
Error with Invalid Validation Data. National Bureau of Economic Research Working
Paper #19928.
Currie, J. and Gruber J. (1996a). Health Insurance Eligibility, Utilization of Medical Care, and
Child Health. Quarterly Journal of Economics, 111, 431-466.
Currie, J. and Gruber, J. (1996b). Saving Babies: The Efficacy and Cost of Recent Changes in
the Medicaid Eligibility of Pregnant Women. Journal of Political Economy, 104, 1263
1296.
98

Dafny, L., & Gruber, J. (2005). Public Insurance and Child Hospitalizations: Access and
Efficient Effects. Journal of Public Economics, 89, 109-129.
Danaei, G., Ding, E. L., Mozaffarian, D., Taylor, B., Rehm, J., Murray, C. J. L., & Ezzati, M.
(2009). The Preventable Causes of Death in the United States: Comparative Risk
Assessment of Dietary, Lifestyle, and Metabolic Risk Factors. PLoS Med, 6(4).
Dave, D. and Kaestner, R. (2009). Health Insurance and Ex Ante Moral Hazard: Evidence from
Medicare. International Journal of Health Care Finance and Economics, 9, 367-390.
Dee, T. S. (1999). State Alcohol Policies, Teen Drinking and Traffic Fatalities. Journal of Public
Economics, 72, 289-315.
DeLeire, T., Dague, L., Leininger, L., Voskuil, K., & Friedsam, D. (2013). Wisconsin
Experience Indicates that Expanding Public Insurance to Low-Income Childless Adults
has Health Care Impacts. Health Affairs, 32(6), 1037-1045.
DeNavas-Walt, C., Proctor, B. D., and Smith, J. C. (2010). Income, Poverty, and Health
Insurance Coverage in the United States: 2009. U.S. Census Bureau. Washington, D.C.:
U.S. Government Printing Office, 1-79.
DeNavas-Walt, C., Proctor, B. D., & Smith, J. C. (2011). Income, Poverty, and Health Insurance
Coverage in the United States: 2010. Current Population Reports Consumer Income (60239).
DeSalvo, K. B., Bloser, N., Reynolds, K., He, J., and Muntner, P. (2006). Mortality Prediction
with a Single General Self-Rated Health Question. A Meta-Analysis. Journal of General
Internal Medicine, 21, 267275.
Dunn, R. A., & Tefft, N. W. (2013). Has Increased Body Weight Made Driving Safer? Health
Economics, Early View.
Ehrlick, I. and Becker, G. S. (1972). Market Insurance, Self-Insurance, and Self-Protection.
Journal of Political Economy, 80, 623-648.
Finkelstein, A. (2007). The Aggregate Effects of Health Insurance: Evidence from the
Introduction of Medicare. Quarterly Journal of Economics, 122(1), 1-37.
Finkelstein, A.N, and McKnight R. (2008). What Did Medicare Do? The Initial Impact of
Medicare on Mortality and Out of Pocket Medical Spending. Journal of Public
Economics, 92, 1644-1668.
Finkelstein, A. N., Taubman, S., Wright, B., Bernstein, M., Gruber, J., Newhouse, J. P., Allen,
H., Baicker, K., and The Oregon Health Study Group. (2012). The Oregon Health
Insurance Experiment: Evidence from the First Year. Quarterly Journal of Economics,
127, 1057-1106.
99

Frijters, P., Haisken-DeNew, J. P., and Shields, M. (2005). The Causal Effect of Income on
Health: Evidence from German Reunification. Journal of Health Economics, 24, 9971017.
Gallet, C. A. (2007). The Demand for Alcohol: a Meta-analysis of Elasticities. Australian Journal
of Agricultural and Resource Economics, 51, 121-135.
Ginde, A. A., Lowe, R. A., & Wiler, J. L. (2012). Health Insurance Status Change and
Emergency Department Use among US Adults. Archives of Internal Medicine, 172(8),
642-647.
Grabowski, D. C., & Morrisey, M. A. (2004). Gasoline Prices and Motor Vehicle Fatalities.
Journal of Policy Analysis and Management, 23(3), 575-593.
Grabowski, D. C., & Morrisey, M. A. (2006). Do Higher Gasoline Taxes Save Lives? Economics
Letters, 90(1), 51-55.
Grossman, M. (1972). On the Concept of Health Capital and Demand for Health. Journal of
Political Economy, 80, 223-255.
Gruber, J. (2011). The Impacts of the Affordable Care Act: How Reasonable Are the Projections?
National Bureau of Economic Research Working Paper #17168.
Hernandes-Boussard, T., Burns, C. S., Wang, N. E., Baker, L. C., & Goldstein, B. A. (2014). The
Affordable Care Act Reduces Emergency Department Use by Young Adults: Evidence
from Three States. Health Affairs, 33(9), 1648-1654.
Idler, E. L. and Benyamini, Y. (1997). Self-Rated Health and Mortality: a Review of TwentySeven Community Studies. Journal of Health and Social Behavior, 38, 21-37.
Jianakoplos, N.A. and Bernasek, A. (1998). Are Women More Risk Averse? Economic Inquiry,
36, 620-630.
Kolstad, J.T. and Kowalski, A.E. (2012). The Impact of Health Care Reform on Hospital and
Preventive Care: Evidence from Massachusetts. Journal of Public Economics, 96, 909929.
Lakdawalla, D. and Philipson, T. (2002). The Growth of Obesity and Technological Change: A
Theoretical and Empirical Investigation. National Bureau of Economic Research
Working Paper # 8946.
Levine, P. B., McKnight R., and Heep S. (2011). How Effective are Public Policies to Increase
Health Insurance Coverage among Young Adults? American Economic Journal:
Economic Policy, 3 (1): 129-156.

100

Lichtenberg, F.R. (2002). The Effects of Medicare on Health Care Utilization and Outcomes. In
Frontiers in Health Policy Research, Volume 5, Cambridge, MA: MIT Press.
Lindahl, M. (2005). Estimating the Effect of Income on Health and Mortality Using
Lottery Prizes as an Exogenous Source of Variation in Income. Journal of Human
Resources, 40(1), 144-168.
McGinnis, M. J., & Foege, W. H. (1993). Actual Causes of Death in the United States. Journal of
the American Medical Association, 270(18), 2207-2212.
Manning, W. G., Newhouse, J. P., Duan, N., Keeler, E. B., Leibowitz, A., and Marquis, M.S.
(1987). Health Insurance and the Demand for Medical Care: Evidence from a
Randomized Experiment. American Economic Review, 77(3), 251-277.
Marton, J. The Impact of the Introduction of Premiums into a SCHIP Program, Journal of Policy
Analysis and Management 26 (2), Spring 2007, pp. 261-279.
Marton, J. and Talbert, J. CHIP Premiums, Health Status, and the Insurance Coverage of
Children, Inquiry 47(3), Fall 2010, pp. 199-214.
Meara, E., Golberstein, E., Zaha, R., Greenfield, S. F., Beardslee, W. R., & Busch, S. H. (2014).
Use of Hospital-Based Services among Young Adults with Behavioral Health Diagnoses
before and after Health Insurance Expansions. JAMA Psychiatry, 71(4), 404-411.
Merluzzi, T. V., & Nairn, R. C. (1999). Adulthood and Aging: Transitions in Health and Health
Cognition. LifeSpan Perspective on Health and Illness, 189-206.
Miller, S. (2011). The Effect of Insurance on Outpatient Emergency Room Visits: An Analysis
of the 2006 Massachusetts Health Reform. Journal of Public Economics, 96, 893-908.
Miller, S. (2012). The Impact of the Massachusetts Health Care Reform on Health Care Use
among Children. American Economic Review: Papers and Proceedings, 102(3), 502-507.
Mokdad, A. H., Marks, J. S., Stroup, D. F., & Gerberding, J. L. (2004). Actual Causes of Death
in the United States, 2000. Journal of the American Medical Association, 291(10), 12381245.
Mokdad, A. H., Marks, J. S., Stroup, D. F., & Gerberding, J. L. (2005). Correction: Actual
Causes of Death in the United States, 2000. Journal of the American Medical Association,
293(3), 293.
Monheit, A., Cantor, J., DeLia, D. and Belloff, D. (2011). How Have State Policies to Expand
Dependent Coverage Affected the Health Insurance Status of Young Adults? Health
Services Research, 46, 251-267.

101

Mulcahy, A., Harris, K., Finegold, K., Kellermann, A., Edelman, L., & Sommers, B. D. (2013).
Insurance Coverage of Emergency Care for Young Adults under Health Reform. New
England Journal of Medicine, 368(22), 2105-2112.
National Conference of State Legislatures. (2010). "Covering Young Adults through their
Parents' or Guardians' Health Policy." Retrieved from
http://www.ncsl.org/research/health/dependent-health-coverage-stateimplementation.aspx
National Highway Traffic Safety Administration (2014). Traffic Safety Facts 2012: AlcoholImpaired Driving. Available at http://www-nrd.nhtsa.dot.gov/Pubs/811870.pdf, accessed
12/14/14.
National Institute on Alcohol Abuse and Alcoholism (2013). College Drinking. Available at
http://pubs.niaaa.nih.gov/publications/CollegeFactSheet/CollegeFactSheet.pdf, accessed
12/14/14.
Nicholson, J. L., Collins, S. R., Mahato, B., Gould, E., Schoen, C., & Rustgi, S. (2009). Rite of
Passage? Why Young Adults Become Uninsured and How New Policies Can Help, 2009
Update. Commonwealth Fund (Issue Brief).
Phillips A. C., Der, G., and Carroll, D. (2010). Self-Reported Health, Self-Reported Fitness, and
all-Cause Mortality: Prospective Cohort Study. British Journal of Health Psychology, 15,
337-346.
Ruhm, C. J. (1996). Alcohol Policies and Highway Vehicle Fatalities. Journal of Health
Economics, 15(4), 435-454.
Ruhm, C. J. (2000). Are Recessions Good For Your Health? Quarterly Journal of Economics,
115(2), 617-650.
Ruhm, C. J. (2002). Does Drinking Really Decrease in Bad Times? Journal of Health Economics,
21(4), 659-678.
Ruhm, C. J. (2005). Healthy Living In Hard Times. Journal of Health Economics, 24(2), 341363.
Ruhm, C. J. (2013). Recessions, Healthy No More? National Bureau of Economic Research
Working Paper #19287.
Saffer, H., & Grossman, M. (1987). Beer Taxes, the Legal Drinking Age, and Young Motor
Vehicle Fatalities. Journal of Legal Studies, 16(2), 351-374.
Schmeiser, M. D. (2009). Expanding Wallets and Waistlines: The Impact of Family
Income on the BMI of Women and Men Eligible for the Earned Income Tax Credit.
Health Economics, 18, 1277-1294.
102

Simmons, W. O., & Zlatoper, T. J. (2010). Obesity and Motor Vehicle Deaths in the USA: a
State-level Analysis. Journal of Economic Studies, 37, 544-556.
Slusky, D. J. G. (2013). Revisiting the Labor Effects of the Affordable Care Act. Princeton
University Working Paper.
Snyder, S. E. and Evans, W. N. (2006). The Impact of Income on Mortality: Evidence from the
Social Security Notch. Review of Economics and Statistics, 88(3), 482-495.
Sommers, B. D., Baicker, K., and Epstein, A. M. (2012). Mortality and Access to Care among
Adults after State Medicaid Expansions. New England Journal of Medicine, 367, 10251034.
Sommers, B. D., Buchmueller, T., Decker, S. L., Carey, C. and Kronick, R. (2013). The
Affordable Care Act Has Led to Significant Gains in Health Insurance and Access to
Care for Young Adults. Health Affairs, 32(1), 165-174.
Sommers, B. D., and Kronick, R. (2012). The Affordable Care Act and Insurance Coverage for
Young Adults. Journal of the American Medical Association, 307 (9): 9134.
Sommers, B. D., Long, S.K., and Baicker, K. (2014). Changes in Mortality After Massachusetts
Health Care Reform: A Quasi-Experimental Study. Annals of Internal Medicine, 160(9),
585-593.
Taubman, S.L., Allen, H.L., Wright, B.J., Baicker, K., and Finkelstein, A.N. (2014). Medicaid
Increases Emergency-Department Use: Evidence from Oregons Health Insurance
Experiment. Science, 343, 263-268.
Tekin, E., McClellan, C., and Minyard, K.J. (2013). Health and Health Behaviors during the
Worst of Times: Evidence from the Great Recession. National Bureau of Economic
Research Working Paper #19234.
Van der Wees, P.J., Zaslavsky, A.M., and Ayanian, J.Z. (2013). Improvements in Health Status
after Massachusetts Health Care Reform. Milbank Quarterly, 91, 663-689.

103

VITA
Yanling Qi is a native of Tianjin, China. Prior to her graduate studies at Georgia State
University, she studied at the Business School of Nankai University, China, where she earned
her B.B.A. in Human Resource Management. As a student there, she also conducted an internal
grant-funded project as a PI.
Yanling began the doctoral program at the Andrew Young School of Policy Studies in
2010. During the first three years, she worked as a GRA for Dr. Paul Kagundu, Dr. Jon
Mansfield, and Dr. James Marton. Her main fields of research interests are health economics,
labor economics, and public economics. She also has research interests in behavioral and
experimental economics. She has presented her work at numerous conferences and seminar
series.
Yanling was the sole instructor for the course of Principles of Microeconomics at
Georgia State, and earned Excellence in College Teaching Certificate and Excellence in
Teaching Economics Award. Besides, she was also awarded the Andrew Young School
dissertation fellowship and the Federal Reserve Bank fellowship.
Starting from 2013, Yanling worked as a research intern and later a visiting scholar in the
Research Department at the Federal Reserve Bank of Atlanta. She mainly worked and
coauthored with Senior Policy Adviser Dr. Julie Hotchkiss on several labor-related policy
projects.
Yanling was awarded her Ph.D. degree in Economics from Georgia State University in
August 2015. She has accepted a tenure-track position at California State University, Long
Beach.

104